[go: up one dir, main page]

WO2024233969A1 - Systems and methods for high-throughput analysis for graphical data - Google Patents

Systems and methods for high-throughput analysis for graphical data Download PDF

Info

Publication number
WO2024233969A1
WO2024233969A1 PCT/US2024/028956 US2024028956W WO2024233969A1 WO 2024233969 A1 WO2024233969 A1 WO 2024233969A1 US 2024028956 W US2024028956 W US 2024028956W WO 2024233969 A1 WO2024233969 A1 WO 2024233969A1
Authority
WO
WIPO (PCT)
Prior art keywords
graphical data
sequence
resolution
scans
scan
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
PCT/US2024/028956
Other languages
French (fr)
Inventor
Vishwa Sanjay PAREKH
Pranav KULKARNI
Adway KANHERE
Paul Yi
Eliot Siegel
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Maryland Baltimore
University of Maryland College Park
Original Assignee
University of Maryland Baltimore
University of Maryland College Park
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Maryland Baltimore, University of Maryland College Park filed Critical University of Maryland Baltimore
Publication of WO2024233969A1 publication Critical patent/WO2024233969A1/en
Anticipated expiration legal-status Critical
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H30/00—ICT specially adapted for the handling or processing of medical images
    • G16H30/40—ICT specially adapted for the handling or processing of medical images for processing medical images, e.g. editing
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
    • G—PHYSICS
    • G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
    • G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
    • G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
    • G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V2201/00—Indexing scheme relating to image or video recognition or understanding
    • G06V2201/03—Recognition of patterns in medical or anatomical images

Definitions

  • AI artificial intelligence
  • communication of graphical data communication of graphical data
  • decoding of graphical data or a combination thereof.
  • BACKGROUND [0003]
  • the utility of artificial intelligence (AI) systems in medical imaging has grown rapidly due in part to the curation of large-scale datasets containing two-dimensional (2D) images and three-dimensional (3D) volumes across various image modalities.
  • AI systems have the ability to learn complex relationships, such as between different tissue types to provide the capability to detect and/or predict the presence of disease.
  • entities have begun to develop and deploy intelligent imaging solutions in hospital systems across a series of different clinical applications, such as Amazon’s HealthLake and Google’s Medical Imaging Suite.
  • media streaming entities require transferring hundreds of gigabytes of data to each viewer device in order to deliver a high-resolution (e.g., 4K) video stream.
  • a high-resolution e.g. 4K
  • using conventional encoding techniques results in greater infrastructure costs, such as due to bandwidth and storage constraints.
  • using conventional encoding techniques results in an inferior experience for the viewer, such as due to an increased time required to decode a transmitted encoding and display each video frame.
  • the present disclosure addresses the shortcomings disclosed above by providing systems and methods for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof via a communication network. More particularly, the systems and methods of the present disclosure train a feature extraction model using a first plurality of graphical data, encode a second one or more graphical data forming an encoded byte stream, match a desired performance to a first training resolution, communicate an optimal subset of an encoded byte stream, terminate communication of the encoded byte stream, or a combination thereof, which allows for less time to complete a process such as two-dimensional (2D) graphical data, three-dimensional (3D) graphical data, four-dimensional (4D) graphical data, or a combination thereof, such as for application with the feature extraction model to complete tasks of classification and/or segmentation across different subject matter (e.g., different organs, tissues, bones, etc.) and imaging modalities.
  • 2D two-dimensional
  • 3D three-dimensional
  • 4D four-dimensional
  • the systems and methods of the present disclosure encode the second one or more graphical data based on a plurality of characteristics that, collectively, define the second one or more graphical, which allows for course-grain utilizations with the second one or more graphical data. In some embodiments, the systems and methods of the present disclosure encode the second one or more graphical data based on a subset of characteristics in the plurality of characteristics that is specific to a respective graphical data in the second one or more graphical data, which allows for a fine-grain utilizations with the second one or more graphical data set.
  • the systems and methods of the present disclosure result in faster turnaround times when processing a request for an evaluation of various graphical data, and reduced overall cost of graphical data storage and communication, without negatively impacting decision making using computational models and one or more computer systems.
  • the system and methods of the present disclosure are utilized match an optimal subset of the encoded byte stream in the form a sequence scan in a plurality of sequence scans of the encoded byte stream that is required for utilization with computer-implemented models, communicate the optimal subset of the encoded byte stream DB1/ 147040794.2 3 Attorney Ref.
  • the systems and methods of the present disclosure enable an end-user, such as a clinician at a second computer system, which includes radiologists, pathologists, oncologists, or the like, to receive a similarly optimized subset of the encoded byte stream, such as for utilizing with high-throughput clinical decision making and diagnosis.
  • the systems and methods of the present disclosure match an optimal resolution for a plurality of graphical data of an encoded byte stream based on, for instance, a clinical use case and/or a form factor (e.g., hardware specifications) of a device including one or more feature extraction models configured to perform an evaluation on a respective modality of graphical data.
  • a form factor e.g., hardware specifications
  • the systems and methods of the present disclosure yield higher throughput, without negatively impacting clinical decision making as well as performance when using a respective computational model with the encoded byte stream, resulting in faster turnaround times, and reduced overall cost of data storage and transmission in comparison to conventional progressive encoding and/or decoding techniques.
  • one aspect of the present disclosure is directed to providing a method for optimizing decoding of graphical data.
  • the method includes using a first computer system.
  • the first computer system includes a processor and memory.
  • the method includes obtaining, in electronic form, a training data set.
  • the training data set includes a first plurality of graphical data.
  • the first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data.
  • the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data.
  • the method includes using the first plurality of graphical data to train a feature extraction model.
  • the method obtains a trained feature extraction model and performance data.
  • the performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances.
  • the method includes receiving a request to evaluation a second one or more graphical data at a second computer system. Moreover, the second one or more graphical data is defined by a second plurality of DB1/ 147040794.2 4 Attorney Ref. No.: 115834-5037-WO characteristics. The second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data.
  • the second plurality of characteristics includes a desired threshold performance.
  • the method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans.
  • each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans.
  • the method includes matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data.
  • the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream at the second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, thereby optimizing decoding of graphical.
  • the first plurality of graphical data includes one or more images taken of a first biological sample.
  • the second one or more graphical data includes one or more images taken of a second biological sample different than the first biological sample.
  • the first plurality of graphical data includes as 2D graphical data, 3D graphical data, 4D graphical data, or a combination thereof.
  • the second one or more graphical data includes 3D graphical data.
  • each scan in the plurality of sequence scans of the 3D graphical data includes a respective 2D layer of the 3D graphical data.
  • the first plurality of graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof.
  • the first plurality of graphical data includes 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images.
  • the second one or more graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non- fungible token assets, or a combination thereof.
  • the sampling resolution of the first plurality of graphical data includes the minimum resolution associated with the first plurality of graphical data.
  • the first modality includes a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof.
  • CT computer tomography
  • MRI magnetic resonance imaging
  • PET positron emission tomography
  • SPE single-photon emission
  • sonography modality or a combination thereof.
  • the first codec includes a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof.
  • the first codec is selected by the first computer system.
  • the plurality of sequence scans includes between 2 scans and 100 scans. [0025] In some embodiments, the plurality of sequence scans includes N scans, in which In some embodiments, M is a number of digital assets associated with the second one or more graphical data. In some embodiments, ⁇ ⁇ ⁇ ⁇ ⁇ 1,2, .. , ⁇ is a native resolution associated with the second one or more graphical data. Moreover, in some embodiments, ⁇ ⁇ ⁇ ⁇ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans.
  • DB1/ 147040794.2 6 Attorney Ref.
  • the first plurality of characteristics further includes one or more ground truth labels with respect to the first modality for the first plurality of graphical data.
  • the using the first plurality of graphical data further includes training the feature extraction model against the one or more ground truth labels.
  • the second plurality of characteristics includes a sampling resolution of the second one or more graphical data.
  • the plurality of sequence scans includes N scans that is based, at least in part, on the sampling resolution of the second one or more graphical data.
  • the second plurality of characteristics further includes a respective capacity of the communication network.
  • the second plurality of characteristics further includes a respective capacity of the first computer system.
  • the feature extraction model includes a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof.
  • the feature extraction model incudes a neutral network model, a support machine mode, a Na ⁇ ve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph-based model, or a combination thereof.
  • the first sequence scan includes less than 5% of the encoded byte stream.
  • the first sequence scan includes less than 5% of the final terminal scan.
  • the desired threshold performance includes a threshold inter- image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation.
  • the encoded byte stream is associated with a first file size of the final terminal scan in the plurality of sequence scans.
  • a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan.
  • the computer system includes one or more processors, a controller, and at least one program is non-transiently stored in the controller and executable by the controller.
  • the at least one program cause the controller to perform a method.
  • the method includes obtaining, in electronic form, a training data set.
  • the training data set includes a first plurality of graphical data.
  • the first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data.
  • the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data.
  • the method includes using the first plurality of graphical data to train a feature extraction model.
  • the method obtains a trained feature extraction model and performance data.
  • the performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances.
  • the method includes receiving a request to evaluation a second one or more graphical data at a second computer system. Moreover, the second one or more graphical data is defined by a second plurality of characteristics. The second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data. Furthermore, the second plurality of characteristics includes a desired threshold performance.
  • the method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. Furthermore, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans.
  • the method includes matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data. Additionally, the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream and one or more instructions to terminate DB1/ 147040794.2 8 Attorney Ref. No.: 115834-5037-WO decoding of the encoded byte stream at the second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, thereby optimizing decoding of graphical.
  • Yet another aspect of the present disclosure is directed to providing a non-transitory computer readable storage medium storing one or more programs.
  • the one or more programs include instructions, which when executed by a computer system, cause the computer system to perform a method.
  • the method includes obtaining, in electronic form, a training data set.
  • the training data set includes a first plurality of graphical data.
  • the first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data.
  • the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data.
  • the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data.
  • the performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances.
  • the method includes receiving a request to evaluation a second one or more graphical data at a second computer system. Moreover, the second one or more graphical data is defined by a second plurality of characteristics. The second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data. Furthermore, the second plurality of characteristics includes a desired threshold performance. The method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data.
  • the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution.
  • Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans.
  • each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans.
  • the method includes matching the desired DB1/ 147040794.2 9 Attorney Ref.
  • the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream at the second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, thereby optimizing decoding of graphical.
  • Yet another aspect of the present disclosure is directed to providing a method for optimizing encoding of graphical data.
  • the method includes using a first computer system.
  • the first computer system includes a processor and memory.
  • the method includes obtaining, in electronic form, a training data set.
  • the training data set includes a first plurality of graphical data.
  • the first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data.
  • the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data.
  • the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data.
  • the performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution.
  • each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances.
  • the method includes receiving a request to evaluation a second one or more graphical data at a second computer system.
  • the second one or more graphical data is defined by a second plurality of characteristics.
  • the second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data.
  • the second plurality of characteristics includes a desired threshold performance.
  • the method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution.
  • each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. Furthermore, each sequence scan in the sequence of the plurality of sequence scans has a DB1/ 147040794.2 10 Attorney Ref. No.: 115834-5037-WO corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans.
  • the first plurality of graphical data includes one or more images taken of a first biological sample.
  • the second one or more graphical data includes one or more images taken of a second biological sample different than the first biological sample.
  • the first plurality of graphical data includes as 2D graphical data, 3D graphical data, 4D graphical data, or a combination thereof.
  • the second one or more graphical data includes 3D graphical data.
  • each scan in the plurality of sequence scans of the 3D graphical data includes a respective 2D layer of the 3D graphical data.
  • the first plurality of graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof.
  • the first plurality of graphical data includes 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images.
  • the second one or more graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non- fungible token assets, or a combination thereof.
  • the sampling resolution of the first plurality of graphical data includes the minimum resolution associated with the first plurality of graphical data.
  • the first modality includes a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof.
  • CT computer tomography
  • MRI magnetic resonance imaging
  • PET positron emission tomography
  • SPE single-photon emission
  • sonography modality or a combination thereof.
  • the first codec includes a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof.
  • the first codec is selected by the first computer system.
  • the plurality of sequence scans includes between 2 scans and 100 scans.
  • the plurality of sequence scans includes N scans, in which In some embodiments, M is a number of digital assets associated with the second one or more graphical data.
  • M is a number of digital assets associated with the second one or more graphical data.
  • ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ 1,2, .. , ⁇ is a native resolution associated with the second one or more graphical data.
  • ⁇ ⁇ ⁇ ⁇ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans.
  • the first plurality of characteristics further includes one or more ground truth labels with respect to the first modality for the first plurality of graphical data.
  • the using the first plurality of graphical data further includes training the feature extraction model against the one or more ground truth labels.
  • the second plurality of characteristics includes a sampling resolution of the second one or more graphical data.
  • the plurality of sequence scans includes N scans that is based, at least in part, on the sampling resolution of the second one or more graphical data.
  • the second plurality of characteristics further includes a respective capacity of the communication network.
  • the second plurality of characteristics further includes a respective capacity of the first computer system. DB1/ 147040794.2 12 Attorney Ref.
  • the feature extraction model includes a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof.
  • the feature extraction model incudes a neutral network model, a support machine mode, a Na ⁇ ve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph-based model, or a combination thereof.
  • the desired threshold performance includes a threshold inter- image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation.
  • the encoded byte stream is associated with a first file size of the final terminal scan in the plurality of sequence scans. Moreover, a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan.
  • Another aspect of the present disclosure is directed to providing computer system.
  • the computer system includes one or more processors, a controller, and at least one program is non-transiently stored in the controller and executable by the controller.
  • the at least one program cause the controller to perform a method.
  • the method includes obtaining, in electronic form, a training data set.
  • the training data set includes a first plurality of graphical data.
  • the first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data. Moreover, the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data.
  • the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data.
  • the performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances.
  • the method includes receiving a request to evaluation a second one or more graphical data at a second computer system.
  • the second one or more graphical data is defined by a second plurality of characteristics.
  • the second plurality of characteristics includes the first modality associated DB1/ 147040794.2 13 Attorney Ref. No.: 115834-5037-WO with a capture of the second one or more graphical data.
  • the second plurality of characteristics includes a desired threshold performance.
  • the method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution.
  • Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. Furthermore, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans.
  • Yet another aspect of the present disclosure is directed to providing a non-transitory computer readable storage medium storing one or more programs.
  • the one or more programs include instructions, which when executed by a computer system, cause the computer system to perform a method.
  • the method includes obtaining, in electronic form, a training data set.
  • the training data set includes a first plurality of graphical data.
  • the first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data.
  • the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data.
  • the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data.
  • the performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution.
  • each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances.
  • the method includes receiving a request to evaluation a second one or more graphical data at a second computer system.
  • the second one or more graphical data is defined by a second plurality of characteristics.
  • the second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data.
  • the second plurality of characteristics includes a desired threshold performance.
  • the method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this DB1/ 147040794.2 14 Attorney Ref.
  • the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution.
  • Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans.
  • each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans.
  • Yet another aspect of the present disclosure is directed to providing a method of optimizing communication of graphical data. The method includes using a first computer system.
  • the first computer system includes a processor and memory.
  • the method includes obtaining, in electronic form, a training data set.
  • the training data set includes a first plurality of graphical data.
  • the first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data.
  • the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data.
  • the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data.
  • the performance data includes a plurality of training resolutions.
  • Each respective training resolution in the plurality of training resolutions is less than the sampling resolution.
  • each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances.
  • the method includes receiving a request to evaluation a second one or more graphical data at a second computer system.
  • the second one or more graphical data is defined by a second plurality of characteristics.
  • the second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data.
  • the second plurality of characteristics includes a desired threshold performance.
  • the method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data.
  • the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution.
  • Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans.
  • DB1/ 147040794.2 15 Attorney Ref. No.: 115834-5037-WO
  • each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans.
  • the method includes matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data. Additionally, the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream. Moreover, the method includes terminating, in accordance with a determination that decoding of the encoded byte stream at the second computer system satisfies a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, the communicating of the encoded byte stream, thereby optimizing communication of graphical data. [0067] In some embodiments, the first plurality of graphical data includes one or more images taken of a first biological sample.
  • the second one or more graphical data includes one or more images taken of a second biological sample different than the first biological sample.
  • the first plurality of graphical data includes as 2D graphical data, 3D graphical data, 4D graphical data, or a combination thereof.
  • the second one or more graphical data includes 3D graphical data.
  • each scan in the plurality of sequence scans of the 3D graphical data includes a respective 2D layer of the 3D graphical data.
  • the first plurality of graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof.
  • the first plurality of graphical data includes 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images.
  • the second one or more graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non- fungible token assets, or a combination thereof.
  • DB1/ 147040794.2 16 Attorney Ref. No.: 115834-5037-WO In some embodiments, the sampling resolution of the first plurality of graphical data includes the minimum resolution associated with the first plurality of graphical data.
  • the first modality includes a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof.
  • CT computer tomography
  • MRI magnetic resonance imaging
  • PET positron emission tomography
  • SPE single-photon emission
  • sonography modality or a combination thereof.
  • the first codec includes a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof.
  • the first codec is selected by the first computer system.
  • the plurality of sequence scans includes between 2 scans and 100 scans. [0079] In some embodiments, the plurality of sequence scans includes N scans, in which In some embodiments, M is a number of digital assets associated with the second one or more graphical data. In some embodiments, ⁇ ⁇ ⁇ ⁇ ⁇ 1,2, .. , ⁇ is a native resolution associated with the second one or more graphical data. Moreover, in some embodiments, ⁇ ⁇ ⁇ ⁇ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans.
  • the first plurality of characteristics further includes one or more ground truth labels with respect to the first modality for the first plurality of graphical data.
  • the using the first plurality of graphical data further includes training the feature extraction model against the one or more ground truth labels.
  • the second plurality of characteristics includes a sampling resolution of the second one or more graphical data.
  • the DB1/ 147040794.2 17 Attorney Ref. No.: 115834-5037-WO plurality of sequence scans includes N scans that is based, at least in part, on the sampling resolution of the second one or more graphical data.
  • the second plurality of characteristics further includes a respective capacity of the communication network.
  • the second plurality of characteristics further includes a respective capacity of the first computer system.
  • the feature extraction model includes a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof.
  • the feature extraction model incudes a neutral network model, a support machine mode, a Na ⁇ ve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph-based model, or a combination thereof.
  • the first sequence scan includes less than 5% of the encoded byte stream.
  • the first sequence scan includes less than 5% of the final terminal scan.
  • the desired threshold performance includes a threshold inter- image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation.
  • the encoded byte stream is associated with a first file size of the final terminal scan in the plurality of sequence scans. Moreover, a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan.
  • Another aspect of the present disclosure is directed to providing computer system.
  • the computer system includes one or more processors, a controller, and at least one program is non-transiently stored in the controller and executable by the controller. The at least one program cause the controller to perform a method.
  • the method includes obtaining, in electronic form, a training data set.
  • the training data set includes a first plurality of graphical data.
  • the first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data.
  • the first plurality of characteristics includes a first modality DB1/ 147040794.2 18 Attorney Ref. No.: 115834-5037-WO associated with a capture of the first plurality of graphical data.
  • the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data.
  • the performance data includes a plurality of training resolutions.
  • Each respective training resolution in the plurality of training resolutions is less than the sampling resolution.
  • each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances.
  • the method includes receiving a request to evaluation a second one or more graphical data at a second computer system.
  • the second one or more graphical data is defined by a second plurality of characteristics.
  • the second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data.
  • the second plurality of characteristics includes a desired threshold performance.
  • the method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data.
  • the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution.
  • Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans.
  • each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans.
  • the method includes matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data.
  • the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream. Moreover, the method includes terminating, in accordance with a determination that decoding of the encoded byte stream at the second computer system satisfies a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, the communicating of the encoded byte stream, thereby optimizing communication of graphical data.
  • Yet another aspect of the present disclosure is directed to providing a non-transitory computer readable storage medium storing one or more programs.
  • the one or more programs include instructions, which when executed by a computer system, cause the DB1/ 147040794.2 19 Attorney Ref.
  • the method includes obtaining, in electronic form, a training data set.
  • the training data set includes a first plurality of graphical data.
  • the first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data.
  • the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data.
  • the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data.
  • the performance data includes a plurality of training resolutions.
  • Each respective training resolution in the plurality of training resolutions is less than the sampling resolution.
  • each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances.
  • the method includes receiving a request to evaluation a second one or more graphical data at a second computer system.
  • the second one or more graphical data is defined by a second plurality of characteristics.
  • the second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data.
  • the second plurality of characteristics includes a desired threshold performance.
  • the method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data.
  • the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution.
  • Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans.
  • each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans.
  • the method includes matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data.
  • the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream. Moreover, the method includes terminating, in accordance with a determination that decoding of the encoded byte stream at the second computer system satisfies a first sequence scan in the plurality of sequence scans that matches or exceeds the DB1/ 147040794.2 20 Attorney Ref. No.: 115834-5037-WO first training resolution, the communicating of the encoded byte stream, thereby optimizing communication of graphical data.
  • Yet another aspect of the present disclosure is directed to providing a method for optimizing decoding of graphical data. The method includes obtaining, in electronic form, a first one or more graphical data.
  • the first one or more graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first one or more of graphical data.
  • the first one or more graphical data is further defined a first modality associated with a capture of the first one or more of graphical data.
  • the method includes forming a byte stream map for the first one or more graphical data.
  • the byte stream map includes a plurality of sequence scans through the sampling resolution.
  • each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans.
  • each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans.
  • the method includes receiving, from a second computer system, a request to evaluate the first one or more graphical data at a first resolution.
  • the first resolution is selected in accordance with a desired threshold performance from performance data.
  • the performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution.
  • Each respective training resolution is associated with a corresponding threshold performance for a trained feature extraction model in a plurality of threshold performances.
  • the method further includes matching the first resolution to a first sequence scan in the plurality of sequence scans in accordance with the byte stream map. Furthermore, the method includes encoding, in accordance with a corresponding encoder of a first codec, the first one or more graphical data, thereby forming an encoded byte stream that includes the plurality of sequence scans through at least the first resolution.
  • the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream at the second computer system when the DB1/ 147040794.2 21 Attorney Ref. No.: 115834-5037-WO decoding decodes the first sequence scan in the plurality of sequence scans, thereby optimizing decoding of graphical data.
  • the first one or more of graphical data includes one or more images taken of a first biological sample.
  • the first one or more graphical data includes one or more images taken of a second biological sample different than the first biological sample.
  • the sampling resolution of the first one or more graphical data includes the minimum resolution associated with the first one or more graphical data.
  • the first modality includes a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof.
  • CT computer tomography
  • MRI magnetic resonance imaging
  • PET positron emission tomography
  • SPE single-photon emission
  • sonography modality or a combination thereof.
  • the first plurality of characteristics further includes one or more ground truth labels with respect to the first modality for the first one or more of graphical data.
  • the first plurality of characteristics further includes a respective capacity of the communication network. [00101] In some embodiments, the first plurality of characteristics further includes a respective capacity of the second computer system. [00102] In some embodiments, the first one or more graphical data includes 3D graphical data and each scan in the plurality of sequence scans of the 3D graphical data includes a respective 2D layer of the 3D graphical data [00103] In some embodiments, the first one or more graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non- fungible token assets, or a combination thereof.
  • the plurality of sequence scans includes between 2 scans and 100 scans. [00105] In some embodiments, the plurality of sequence scans includes N scans, in which In some embodiments, M is a number of digital assets associated with DB1/ 147040794.2 22 Attorney Ref. No.: 115834-5037-WO the second one or more graphical data. In some embodiments, ⁇ ⁇ ⁇ ⁇ ⁇ 1,2, .. , ⁇ is a native resolution associated with the second one or more graphical data. Moreover, in some embodiments, ⁇ ⁇ ⁇ ⁇ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans.
  • the plurality of sequence scans includes N scans, in which ⁇ ⁇ 1, M is a number of digital assets associated with the second one or more graphical data, ⁇ ⁇ ⁇ ⁇ ⁇ 1,2, .. , ⁇ is a native resolution associated with the second one or more graphical data, and ⁇ ⁇ ⁇ ⁇ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans.
  • the first plurality of characteristics includes a sampling resolution of the first one or more graphical data, in which the plurality of sequence scans includes N scans based, at least in part, on the sampling resolution of the first one or more graphical data.
  • the second plurality of graphical data includes two- dimensional (2D) graphical data, three-dimensional (3D) graphical data, four-dimensional (4D) data, or a combination thereof.
  • the second plurality of graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non- fungible token assets, or a combination thereof.
  • the second plurality of graphical data includes 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images.
  • the method prior to the receiving C), the method further includes training the feature extraction model against the one or more ground truth labels.
  • the desired threshold performance includes a threshold inter- image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation.
  • the feature extraction model includes a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof. DB1/ 147040794.2 23 Attorney Ref.
  • the feature extraction model includes a neutral network model, a support machine mode, a Na ⁇ ve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph-based model, or a combination thereof.
  • the first codec includes a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof.
  • the first codec is selected by the second computer system.
  • the first sequence scan includes less than 5% of the encoded byte stream.
  • the first sequence scan includes less than 5% of a final terminal scan.
  • the encoded byte stream is associated with a first file size of the final terminal scan in the plurality of sequence scans, and a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan.
  • Figure 1 illustrates an exemplary system topology including a distributed computer system including a graphical data system and one or more client devices, in accordance with an exemplary embodiment of the present disclosure
  • Figures 2A and 2B collectively illustrates a graphical data system for generating an encoded byte stream, matching a desired threshold performance associated with a request to evaluate graphical data to a first training resolution, communicating the encoded byte stream, DB1/ 147040794.2 24 Attorney Ref.
  • Figure 3 illustrates a client device, in accordance with an embodiment of the present disclosure
  • Figures 4A, 4B, 4C, 4D, and 4E collectively illustrate exemplary methods for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof, in which optional embodiments are indicated by dashed boxes, in accordance with some embodiments of the present disclosure
  • Figures 5A, 5B, 5C, 5D, and 5E collectively illustrate exemplary methods for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof, in which optional embodiments are indicated by dashed boxes, in accordance with some embodiments of the present disclosure
  • Figure 6 illustrates a chart depicting an overview of an architecture for optimizing encoding of graphical data
  • Figure 10 illustrates a chart depicting a comparison of graphical data and model metrics for 3D graphical data applied to a segmentation model for liver and spleen segmentation using a first NifTI codec and a second HTJ2K version of a second plurality of graphical data (e.g., the Multi-Atlas Labelling Beyond the Cranial Vault (BTCV) external test data set), in accordance with some embodiments of the present disclosure;
  • Figure 11 illustrates a representation of a plurality of sequence scans of an encoded byte stream for a second one or more graphical data, in accordance with some embodiments of the present disclosure;
  • Figure 12 illustrates a block diagram of a HTJ2K codec including an encoder and a decoder for communicating an encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream, in accordance with some embodiments of the present disclosure;
  • Figure 13A illustrates
  • FIG. 17A, 17B, and 17C collectively depict charts of mean AUROC scores for 2D classification of abnormalities using chest x-ray graphical data on the original graphical data and across each HTJ2K scan, including an optimal resolution (hatched) for an Intelligent Streaming framework (“ISLE”), using the original graphical data such
  • the present disclosure provides systems and methods for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof are provided.
  • a training data set including first graphical data defined by a sampling resolution and a first modality are obtained.
  • the first graphical data e.g., first plurality of graphical data
  • the performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution of the first graphical data.
  • each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances.
  • the plurality of training resolutions together with the plurality of threshold performances provide for a standard of comparison for future applications with a respective feature extraction model.
  • a request to evaluate second graphical data (e.g., second one or more graphical data) is received.
  • the second graphical data is defined by the first modality and a desired threshold performance.
  • the second graphical data is distinct from the first graphical data.
  • the second graphical data is encoded forming an encoded byte stream including a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of DB1/ 147040794.2 28 Attorney Ref.
  • each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans.
  • the corresponding progressive resolution allows for each respective sequence scan to have a corresponding unique resolution.
  • the desired threshold performance is matched to a first training resolution in the plurality of training resolutions in the performance data.
  • the encoded byte stream is communicated to the second computer system, via a communication network, in accordance with the first codec.
  • the communicating the encoded byte stream includes one or more instructions to terminate decoding of the encoded byte stream at a second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution. Still further, in some embodiments, the one or more instructions is configured to terminate communication of the encoded byte stream at a second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution.
  • the term “if” may be construed to mean “when” or “upon” or “in response to determining” or “in response to detecting,” depending on the context.
  • the phrase “if it is determined” or “if [a stated condition or event] is detected” may be construed to mean “upon determining” or “in response to determining” or “upon detecting [the stated condition or event]” or “in response to detecting [the stated condition or event],” depending on the context.
  • the term “about” or “approximately” can mean within an acceptable error range for the particular value as determined by one of ordinary skill in the art, which can depend in part on how the value is measured or determined, e.g., the limitations of the measurement system. For example, “about” can mean within 1 or more than 1 standard deviation, per the practice in the art. “About” can mean a range of ⁇ 20%, ⁇ 10%, ⁇ 5%, or ⁇ 1% of a given value. Where particular values are described in the application and claims, unless otherwise stated, the term “about” means within an acceptable error range for the particular value. The term “about” can have the meaning as commonly understood by one of ordinary skill in the art.
  • the term “about” can refer to ⁇ 10%.
  • the term “about” can refer to ⁇ 5%.
  • the term “lossless,” means a compression ratio of 0% or about 0%.
  • the term “lossy,” means a compression ratio of less or equal to 5%.
  • the term “scan,” “sequence scan” and “decomposition” are used interchangeably herein unless expressly stated otherwise.
  • the term “AxB,” denotes a resolution of two-dimensional graphical data in which “A” is a number of pixels in a horizontal direction and “B” is a number of pixels in a vertical direction.
  • a model includes an unsupervised learning algorithm.
  • an unsupervised learning algorithm is cluster analysis.
  • a model includes supervised machine learning.
  • Nonlimiting examples of supervised learning algorithms include, but are not limited to, logistic regression, neural networks, support vector machines, Naive Bayes algorithms, nearest neighbor algorithms, random forest algorithms, decision tree algorithms, boosted trees algorithms, multinomial logistic regression algorithms, linear models, linear regression, Gradient Boosting, mixture models, hidden Markov models, Gaussian NB algorithms, linear discriminant analysis, or any combinations thereof.
  • a model is a multinomial classifier algorithm.
  • a model is a 2-stage stochastic gradient descent (SGD) model.
  • a model is a deep neural network (e.g., a deep-and-wide sample-level model). [00162] Neural networks.
  • the model is a neural network (e.g., a convolutional neural network and/or a residual neural network).
  • Neural network algorithms also known as artificial neural networks (ANNs), include convolutional and/or residual neural network algorithms (deep learning algorithms).
  • ANNs artificial neural networks
  • neural networks are machine learning algorithms that are trained to map an input dataset to an output dataset, where the neural network includes an interconnected group of nodes organized into multiple layers of nodes.
  • the neural network architecture includes at least an input layer, one or more hidden layers, and an output layer.
  • the neural network includes any total number of layers, and any number of hidden layers, where the hidden layers function as trainable feature extractors that allow mapping of a set of input data to an output value or set of output values.
  • a deep learning algorithm is a neural network including a plurality of hidden layers, e.g., two or more hidden layers.
  • each layer of the neural network includes a number of nodes (or “neurons”).
  • a node receives input that comes either directly from the input data or the output of nodes in previous layers, and performs a specific operation, e.g., a summation operation.
  • a connection from an input to a node is associated with a parameter (e.g., a weight and/or weighting factor).
  • the node sums up the products of all pairs of inputs, x i , and their associated parameters.
  • the weighted sum is offset with a bias, b.
  • the output of a node or neuron is gated using a threshold or activation function, f, which, in some instances, is a linear or non-linear function.
  • the activation function is, for example, a rectified linear unit (ReLU) DB1/ 147040794.2 32 Attorney Ref.
  • the weighting factors, bias values, and threshold values, or other computational parameters of the neural network are “taught” or “learned” in a training phase using one or more sets of training data.
  • the parameters are trained using the input data from a training dataset and a gradient descent or backward propagation method so that the output value(s) that the ANN computes are consistent with the examples included in the training dataset.
  • the parameters are obtained from a back propagation neural network training process.
  • Any of a variety of neural networks are suitable for use in accordance with the present disclosure. Examples include, but are not limited to, graph neural networks, feedforward neural networks, radial basis function networks, recurrent neural networks, residual neural networks, convolutional neural networks, residual convolutional neural networks, and the like, or any combination thereof.
  • the machine learning makes use of a pre-trained and/or transfer-learned ANN or deep learning architecture.
  • a deep neural network model includes an input layer, a plurality of individually parameterized (e.g., weighted) convolutional layers, and an output scorer.
  • the parameters (e.g., weights) of each of the convolutional layers as well as the input layer contribute to the plurality of parameters (e.g., weights) associated with the deep neural network model.
  • at least 50 parameters, at least 100 parameters, at least 1,000 parameters, at least 2,000 parameters or at least 5,000 parameters are associated with the deep neural network model.
  • deep neural network models require a computer to be used because they cannot be mentally solved.
  • the model output needs to be determined using a computer rather than mentally in such embodiments.
  • Krizhevsky et al. 2012, “Imagenet classification with deep convolutional neural networks,” in Advances in Neural Information Processing Systems 2, Pereira, Burges, Bottou, Weinberger, eds., pp.1097-1105, Curran Associates, Inc.; Zeiler, 2012 “ADADELTA: an adaptive learning rate method,” CoRR, vol. abs/1212.5701; and Rumelhart et al., 1988, “Neurocomputing: Foundations of research,” ch. Learning DB1/ 147040794.2 33 Attorney Ref.
  • Neural network algorithms including convolutional neural network algorithms, suitable for use as models are disclosed in, for example, Vincent et al., 2010, “Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion,” J Mach Learn Res 11, pp.3371-3408; Larochelle et al., 2009, “Exploring strategies for training deep neural networks,””dec J Mach Learn Res 10, pp.1-40; and Hassoun, 1995, Fundamentals of Artificial Neural Networks, Massachusetts Institute of Technology, each of which is hereby incorporated by reference.
  • Additional example neural networks suitable for use as models are disclosed in Duda et al., 2001, Pattern Classification, Second Edition, John Wiley & Sons, Inc., New York; and Hastie et al., 2001, The Elements of Statistical Learning, Springer-Verlag, New York, each of which is hereby incorporated by reference in its entirety. Additional example neural networks suitable for use as models are also described in Draghici, 2003, Data Analysis Tools for DNA Microarrays, Chapman & Hall/CRC; and Mount, 2001, Bioinformatics: sequence and genome analysis, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York, each of which is hereby incorporated by reference in its entirety. [00167] Ensembles of models and boosting.
  • an ensemble (two or more) of models is used.
  • a boosting technique such as AdaBoost is used in conjunction with many other types of learning algorithms to improve the performance of the model.
  • AdaBoost boosting technique
  • the output of any of the models disclosed herein, or their equivalents is combined into a weighted sum that represents the final output of the boosted model.
  • the plurality of outputs from the models is combined using any measure of central tendency known in the art, including but not limited to a mean, median, mode, a weighted mean, weighted median, weighted mode, etc.
  • the plurality of outputs is combined using a voting method.
  • a respective model in the ensemble of models is weighted or unweighted.
  • the term “parameter” refers to any coefficient or, similarly, any value of an internal or external element (e.g., a weight and/or a hyperparameter) in an algorithm, model, regressor, and/or classifier that can affect (e.g., modify, tailor, and/or adjust) one or more inputs, outputs, and/or functions in the algorithm, model, regressor and/or classifier.
  • a parameter refers to any coefficient, weight, and/or hyperparameter that can be used to control, modify, tailor, and/or adjust the behavior, DB1/ 147040794.2 34 Attorney Ref. No.: 115834-5037-WO learning, and/or performance of an algorithm, model, regressor, and/or classifier.
  • a parameter is used to increase or decrease the influence of an input (e.g., a feature) to an algorithm, model, regressor, and/or classifier.
  • a parameter is used to increase or decrease the influence of a node (e.g., of a neural network), where the node includes one or more activation functions.
  • Assignment of parameters to specific inputs, outputs, and/or functions is not limited to any one paradigm for a given algorithm, model, regressor, and/or classifier but can be used in any suitable algorithm, model, regressor, and/or classifier architecture for a desired performance.
  • a parameter has a fixed value.
  • a value of a parameter is manually and/or automatically adjustable.
  • a value of a parameter is modified by a validation and/or training process for an algorithm, model, regressor, and/or classifier (e.g., by error minimization and/or backpropagation methods).
  • an algorithm, model, regressor, and/or classifier of the present disclosure includes a plurality of parameters.
  • the plurality of parameters is n parameters, where: n ⁇ 2; n ⁇ 5; n ⁇ 10; n ⁇ 25; n ⁇ 40; n ⁇ 50; n ⁇ 75; n ⁇ 100; n ⁇ 125; n ⁇ 150; n ⁇ 200; n ⁇ 225; n ⁇ 250; n ⁇ 350; n ⁇ 500; n ⁇ 600; n ⁇ 750; n ⁇ 1,000; n ⁇ 2,000; n ⁇ 4,000; n ⁇ 5,000; n ⁇ 7,500; n ⁇ 10,000; n ⁇ 20,000; n ⁇ 40,000; n ⁇ 75,000; n ⁇ 100,000; n ⁇ 200,000; n ⁇ 500,000, n ⁇ 1 x 10 6 , n ⁇ 5 x 10 6 , or n ⁇ 1 x 10 7 .
  • n is between 10,000 and 1 x 10 7 , between 100,000 and 5 x 10 6 , or between 500,000 and 1 x 10 6 .
  • the algorithms, models, regressors, and/or classifier of the present disclosure operate in a k-dimensional space, where k is a positive integer of 5 or greater (e.g., 5, 6, 7, 8, 9, 10, etc.). As such, the algorithms, models, regressors, and/or classifiers of the present disclosure cannot be mentally performed.
  • the term “untrained model” refers to a machine learning model or algorithm, such as a classifier or a neural network, that has not been trained on a target dataset.
  • “training a model” refers to the process of training an untrained or partially trained model (e.g., “an untrained or partially trained neural network”).
  • the term “untrained model” does not exclude the possibility that transfer learning techniques are used in such training of the untrained or partially trained model.
  • auxiliary training datasets that can be used to complement the primary training dataset in training the untrained model in the present disclosure.
  • two or more auxiliary training datasets, three or more auxiliary training datasets, four or more auxiliary training datasets or five or more auxiliary training datasets are used to complement the primary training dataset through transfer learning, where each such auxiliary dataset is different than the primary training dataset. Any manner of transfer learning is used, in some such embodiments. For instance, consider the case where there is a first auxiliary training dataset and a second auxiliary training dataset in addition to the primary training dataset.
  • the parameters learned from the first auxiliary training dataset (by application of a first model to the first auxiliary training dataset) are applied to the second auxiliary training dataset using transfer learning techniques (e.g., a second model that is the same or different from the first model), which in turn results in a trained intermediate model whose parameters are then applied to the primary training dataset and this, in conjunction with the primary training dataset itself, is applied to the untrained model.
  • transfer learning techniques e.g., a second model that is the same or different from the first model
  • a first set of parameters learned from the first auxiliary training dataset (by application of a first model to the first auxiliary training dataset) and a second set of parameters learned from the second auxiliary training dataset (by application of a second model that is the same or different from the first model to the second auxiliary training dataset) are each individually applied to a separate instance of the primary training dataset (e.g., by separate independent matrix multiplications) and both such applications of the parameters to separate instances of the primary training dataset in conjunction with the primary training dataset itself (or some reduced form of the primary training dataset such as principal components or regression coefficients learned from the primary training set) are then applied to the untrained model in order to train the untrained model.
  • model i refers to the i th model in a plurality of models (e.g., a model 118-i in a plurality of models 118).
  • descriptions of devices and systems will include implementations of one or more computers.
  • descriptions of devices and systems will include implementations of one or more computers.
  • a graphical data system 200 is represented as a single device that includes all the functionality of a computer system.
  • a client device 300 is represented as a single device that includes all the functionality of a computer system.
  • the functionality of the graphical data system 200 is spread across any number of networked computers and/or reside on each of several networked computers and/or by hosted on one or more virtual machines and/or containers at a remote location accessible across a communications network (e.g., communications network 186 of Figure 1).
  • FIG. 1 merely serves to describe the features of an embodiment of the present disclosure in a manner that will be readily understood to one of skill in the art.
  • Figures 2A and 2B collectively depicts a block diagram of a graphical data system 200 according to some embodiments of the present disclosure.
  • the graphical data system 200 at least facilitates optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof, such as by obtaining a training data set (e.g., block 404 of Figure 4A), using a first plurality of graphical data to train a feature extraction model (e.g., block 418 of Figure 4B), receiving a request to evaluate a second one or more graphical data (e.g., block 428 of Figure 4C), encoding (e.g., first sequence scan 1010-1 of Figure 11) an encoded byte stream for communication via a communication network 186 (e.g., block 442 of Figure 4D), matching a desired threshold performance to first training resolution (e.g., block 456 of Figure 4E), communicating the encoded byte stream (e.g., block 458 of Figure 4E), generating one or more instructions (e.g., DB1/ 147040794.2 37 Attorney Ref.
  • a training data set e.g.,
  • the communication network 186 optionally includes the Internet, one or more local area networks (LANs), one or more wide area networks (WANs), other types of networks, or a combination of such networks.
  • Examples of communication networks 186 include the World Wide Web (WWW), an intranet and/or a wireless network, such as a cellular telephone network, a wireless local area network (LAN) and/or a metropolitan area network (MAN), and other devices by wireless communication.
  • WWW World Wide Web
  • LAN wireless local area network
  • MAN metropolitan area network
  • the wireless communication optionally uses any of a plurality of communications standards, protocols and technologies, including Global System for Mobile Communications (GSM), Enhanced Data GSM Environment (EDGE), high-speed downlink packet access (HSDPA), high-speed uplink packet access (HSUPA), Evolution, Data-Only (EV-DO), HSPA, HSPA+, Dual-Cell HSPA (DC-HSPDA), long term evolution (LTE), near field communication (NFC), wideband code division multiple access (W-CDMA), code division multiple access (CDMA), time division multiple access (TDMA), Bluetooth, Wireless Fidelity (Wi-Fi) (e.g., IEEE 802.11a, IEEE 802.11ac, IEEE 802.11ax, IEEE 802.11b, IEEE 802.11g and/or IEEE 802.11n), voice over Internet Protocol (VoIP), Wi- MAX, a protocol for e-mail (e.g., Internet message access protocol (IMAP) and/or post office protocol (POP)), instant messaging (e.g., extensible messaging
  • the graphical data system 200 includes one or more processing units (CPUs) 172, a network or other communications interface 174, and memory 192.
  • the graphical data system 200 includes a user interface 176.
  • the user interface 176 typically includes a display 178 for presenting media, such as a result by a plurality of models (e.g., first model 118-1, second model 118-2, ..., model X 118-X of Figure 2B) or a portion (e.g., some or all) of a plurality of graphical data (e.g., first graphical data 108-1 of Figure 1, second graphical data 108-2 of Figure 11, second graphical data 108- 2 of Figure 12, etc.).
  • a plurality of models e.g., first model 118-1, second model 118-2, ..., model X 118-X of Figure 2B
  • a portion e.g., some or all
  • the display 178 is integrated within the graphical DB1/ 147040794.2 38 Attorney Ref. No.: 115834-5037-WO data system 200 (e.g., housed in the same chassis as the CPU 172 and the memory 192).
  • the graphical data system 200 includes one or more input device(s) 180, which allow a subject to interact with the graphical data system 200.
  • the input devices 180 include a keyboard, a mouse, and/or other input mechanisms.
  • the display 178 includes a touch- sensitive surface (e.g., where display 178 is a touch-sensitive display or the graphical data system 200 includes a touch pad).
  • the graphical data system 200 presents media to a user through the display 178.
  • Examples of media presented by the display 178 include one or more images, a video, audio (e.g., waveforms of an audio sample), or a combination thereof.
  • the one or more images, the video, the audio, or the combination thereof is presented by the display 178 through a client application 130.
  • the audio is presented through an external device (e.g., speakers, headphones, input/output (I/O) subsystem, etc.) that receives audio information from the graphical data system 200 and presents audio data based on this audio information.
  • an external device e.g., speakers, headphones, input/output (I/O) subsystem, etc.
  • the user interface 176 also includes an audio output device, such as speakers or an audio output for connecting with speakers, earphones, or headphones.
  • the memory 192 includes high-speed random access memory, such as DRAM, SRAM, DDR RAM, or other random access solid state memory devices, and optionally also includes non-volatile memory, such as one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, or other non-volatile solid state storage devices.
  • the memory 192 may optionally include one or more storage devices remotely located from the CPU(s) 172.
  • the memory 192, or alternatively the non-volatile memory device(s) within the memory 192 includes a non-transitory computer readable storage medium.
  • the memory 192 can include mass storage that is remotely located with respect to the CPU(s) 172.
  • some data stored in memory 192 may in fact be hosted on devices that are external to the graphical data system 200, but that can be electronically accessed by the graphical data system 200 over an Internet, intranet, or other form of network 186 or electronic cable using communication interface 184.
  • the memory 192 of the graphical data system 200 optimizing decoding of graphical data 108 stores: DB1/ 147040794.2 39 Attorney Ref.
  • an operating system 102 e.g., ANDROID, iOS, DARWIN, RTXC, LINUX, UNIX, OS X, WINDOWS, or an embedded operating system such as VxWorks
  • an electronic address 104 associated with the graphical data system 200 that identifies graphical data system 200 (e.g., within the communication network 186);
  • a graphical data store 106 that stores a record of graphical data (e.g., first plurality of graphical data 108-1, second one or more graphical data 108-2, ..., graphical data C 108- C of Figure 2A), each graphical data 108 is defined by a plurality of characteristics (e.g., first plurality of characteristics 110-1 of Figure 2A), that collectively characterize a corresponding graphical data 108 by one or more characteristics (e.g., value of first characteristic 112-1-1, value of second characteristic 112-2-1,
  • the graphical data system 200 includes an operating system 102 that includes procedures for handling various basic system services.
  • the operating system 102 e.g., iOS, ANDROID, DARWIN, RTXC, LINUX, UNIX, OS X, WINDOWS, or an embedded operating system such as VxWorks
  • the operating system 102 includes various software components and or drivers for controlling and managing general system tasks (e.g., memory management, storage device control, power management, etc.) and facilitates communication between various hardware and software components.
  • an optional electronic address 104 is associated with the graphical data system 200.
  • the optional electronic address 104 is utilized to at least uniquely identify the graphical data system 200 from other devices and components of the distributed system 100, such as other client devices 300 having access to the communications network 186.
  • the electronic address 104 is utilized to receive a request from a client device 300 to communicate a graphical data 108 set through an encoded byte stream 114 in the form of a first codec 124-1.
  • the present disclosure is not limited thereto.
  • the graphical data system 200 includes a graphical data store 106 that stores a variety of graphical data 108, such as one or more graphical data sets (e.g., first plurality of graphical data 108-1, second one or more graphical data 108-2, ..., plurality of graphical data C 108-C of Figure 2A).
  • the graphical data store 106 is configured to retain between three and 20 graphical data 108 sets, between 5 and 40 graphical data sets, between 15 and 100 graphical data sets, or between 25 and 150 graphical data sets.
  • the graphical data store 106 is configured to retain between three and 20 graphical data sets 108, between 5 and 40 graphical data sets, between 15 and 100 graphical data sets, or between 25 and 150 graphical data sets. In some embodiments, the graphical data store 106 is configured to retain at least 5 graphical data sets 108, at least 15 graphical data sets, at least 25 graphical data sets, at least 50 graphical data sets, at least 100 graphical data sets, or at least 150 graphical data sets. In some embodiments, the graphical data store 106 is configured to retain at most 5 graphical data sets 108, at most 15 graphical data sets, at most 25 graphical data sets, at most 50 graphical data sets, at most 100 graphical data sets, or at most 150 graphical data sets.
  • Each respective graphical data 108 is defined by a corresponding plurality of characteristics (e.g., first plurality of characteristics 110-1 define first plurality of graphical data 108-1 of Figure 2A).
  • each respective plurality of graphical data 108 is defined by between three and 20 characteristics 112, between 5 and 40 characteristics, between 15 and 100 characteristics, or between 25 and 150 characteristics.
  • each respective plurality of graphical data 108 is defined by between three and 20 characteristics 112, between 5 and 40 characteristics, between 15 and 100 characteristics, or between 25 and 150 characteristics.
  • each respective plurality of graphical data 108 is defined by at least 5 characteristics 112, at least 15 characteristics, at least 25 characteristics, at least 50 characteristics, at least 100 characteristics, or at least 150 characteristics. In some embodiments, each respective plurality of graphical data 108 is defined by at most 5 characteristics 112, at most 15 characteristics, at most 25 characteristics, at most 50 characteristics, at most 100 characteristics, or at most 150 characteristics 112. [00184] Collectively, the plurality of characteristics 110 characterize the corresponding graphical data 108 by one or more characteristics 112. The one or more characteristics provide information about one or more aspects of the corresponding graphical data 108.
  • the one or more characteristics 112 includes a first characteristic 112-1 associated with a corresponding modality of the plurality of graphical data 108 (e.g., a first value of the first characteristic 112-1 is associated with a first modality, a second value of the first characteristic 112-1 is associated with a second modality different than the first modality, etc.), a second characteristic 112-2 associated with a sampling resolution of the plurality of graphical data 108, a third characteristic 112-3 associated with a file format of the plurality of graphical data, a fourth characteristic 112-4 associated with a subject matter (e.g., content) of the plurality of graphical data, a fifth characteristic 112-5 associated with a capture setting (e.g., camera setting) associated with the plurality of graphical data, or a combination thereof.
  • a first characteristic 112-1 associated with a corresponding modality of the plurality of graphical data 108
  • a second characteristic 112-2 associated with a sampling resolution of the plurality of graphical data 108
  • a respective characteristic 112 in the plurality of characteristics 112 associated with the plurality of graphical data 108 is utilized as input for a model 118, such as for identifying a first resolution in a plurality of resolutions for optimizing decoding of the plurality of graphical data 108 and/or determining if a corresponding desired threshold performance is satisfied.
  • DB1/ 147040794.2 42 Attorney Ref.
  • the third characteristic 112-3 associated with the file format of the plurality of graphical data 108 allows for retaining graphical data in any electronic image file format, including but not limited to JPEG/JFIF, TIFF, Exif, PDF, EPS, GIF, BMP, PNG, PPM, PGM, PBM, PNM, WebP, HDR raster formats, HEIF, BAT, BPG, DEEP, DRW, ECW, FITS, FLIF, ICO, ILBM, IMG, PAM, PCX, PGF, JPEG XR, Layered Image File Format, PLBM, SGI, SID, CD5, CPT, PSD, PSP, XCF, PDN, CGM, SVG, PostScript, PCT, WMF, EMF, SWF, XAML, and/or RAW.
  • the ability of the graphical data system 200 to retain a variety of electronic allows for a respective codec 124 to compress and reconstruct the plurality of graphical data 108 in accordance with a corresponding encoder 126 and a corresponding decoder 128 with more than one file type, such as with two-dimensional (2D) graphical data, three-dimensional (3D) graphical data, four-dimensional (4D) graphical data, or a combination thereof.
  • the plurality of graphical data 108 is obtained in any electronic color mode, including but not limited to grayscale, bitmap, indexed, RGB, CMYK, HSV, lab color, duotone, and/or multichannel.
  • the plurality of graphical data 108 is manipulated (e.g., stitched, compressed and/or flattened).
  • the plurality of graphical data 108 has a file size is between 1 KB and 1 MB, between 1 MB and 0.5 GB, between 0.5 GB and 5 GB, between 5 GB and 10 GB, between 10 GB and 500 GB, or greater.
  • the plurality of graphical data 108 has a file size of at least 1 KB, a file size of at least 1 MB, a file size of at least 0.5 GB, or a file size of at least 10 GB.
  • the plurality of graphical data 108 includes between 1 million and 25 million pixels. In some embodiments, each resolution is represented by five or more, ten or more, 100 or more, 1,000 or more contiguous pixels in an image. In some embodiments, each resolution is represented by between 1,000 and 250,000 contiguous pixels in a native image 125. [00188] In some embodiments, the plurality of graphical data 108 is represented as an array (e.g., matrix) including a plurality of pixels, such that the location of each respective pixel in the plurality of pixels in the array (e.g., matrix) corresponds to its original location in the image.
  • an array e.g., matrix
  • the plurality of graphical data 108 is represented as a vector including a plurality of pixels, such that each respective pixel in the plurality of pixels in the vector includes spatial information corresponding to its original location in the image.
  • a pixel includes one or more pixel values (e.g., an intensity value).
  • each respective pixel in the plurality of pixels includes one pixel intensity value, such that the plurality of pixels represents a single-channel image including a one-dimensional integer vector including the respective pixel values for each respective pixel.
  • an 8-bit single-channel image (e.g., grey-scale) includes 2 8 or 256 different pixel values (e.g., 0-255).
  • each respective pixel in the plurality of pixels of an image includes a plurality of pixel values, such that the plurality of pixels represents a multi-channel image including a multi- dimensional integer vector, where each vector element represents a plurality of pixel values for each respective pixel.
  • a 24-bit 3-channel image e.g., RGB color
  • each vector element includes 3 components, each between 0-255.
  • an n-bit image of the plurality of graphical data includes up to 2 n different pixel values, where n is any positive integer. See, Uchida, 2013, “Image processing and recognition for biological images,” Develop. Growth Differ., 55, pg.523-549, doi:10.1111/dgd.12054, which is hereby incorporated herein by reference in its entirety for all purposes.
  • the graphical data system 200 includes a model library 116 that stores a plurality of models 118 (e.g., classifiers, regressors, clustering, etc.).
  • the model library 116 stores two more models 118 (e.g., a first feature extraction model 118-1 and a second feature extraction model 118-2), three or more models (e.g., a first segmentation model 118-1, a second first classification model 118-2, a third second classification model 118-3), four or more models, ten or more models, 50 or more models, or 100 or more models.
  • a model 118 in the plurality of models 118 of the model library 116 is implemented as an artificial intelligence engine.
  • the model 118 includes one or more feature extraction models including one or more segmentation models 118, classification models 118, one or more gradient boosting models 118, one or more random forest models 118, one or more neural network (NN) models 118, one or more regression models, one or more Na ⁇ ve Bayes models 118, one or more machine learning algorithms (MLA) 118, or a combination thereof.
  • an MLA or a NN is trained from a training data set that includes one or more features identified from a data set.
  • MLAs include supervised algorithms (such as algorithms where the features/classifications in the data set are annotated) using linear regression, logistic regression, decision trees, classification and regression trees, Na ⁇ ve DB1/ 147040794.2 44 Attorney Ref. No.: 115834-5037-WO Bayes, nearest neighbor clustering; unsupervised algorithms (such as algorithms where no features/classification in the data set are annotated a priori), such as means clustering, principal component analysis, random forest, adaptive boosting; and semi-supervised algorithms (such as algorithms where an incomplete number of features/classifications in the data set are annotated) using generative approach (such as a mixture of Gaussian distributions, mixture of multinomial distributions, hidden Markov models), low density separation, graph-based approaches (such as minimum cut, harmonic function, manifold regularization, etc.), heuristic approaches, or support vector machines.
  • generative approach such as a mixture of Gaussian distributions, mixture of multinomial distributions, hidden Markov models
  • graph-based approaches such as
  • Neural network models 118 include conditional random fields models 118, convolutional neural network (CNN) models 118, attention based neural network models 118, deep learning models 118, long short term memory network model 118, or other neural models 118.
  • CNN convolutional neural network
  • MLA and neural networks identify distinct approaches to machine learning, the terms may be used interchangeably herein.
  • a reference to MLA may include a corresponding NN or a reference to NN may include a corresponding MLA unless explicitly stated otherwise.
  • the training of a respective model 118 includes providing one or more optimized data sets, such as a first plurality of graphical data 108-1, labeling features as they occur (e.g., as a characteristic 112 associated with the plurality of graphical data 108), and training the MLA to predict or classify based on new inputs.
  • Artificial NNs are efficient computing models 118 which have shown their strengths in solving hard problems in artificial intelligence. For instance, artificial NNs have also been shown to be universal approximators, that is, they can represent a wide variety of functions when given appropriate parameters. [00193]
  • One of skill in the art will readily appreciate other models 118 that are applicable to the systems and methods of the present disclosure.
  • the systems and methods of the present disclosure utilize more than one model 118 to provide an evaluation (e.g., arrive at an evaluation given one or more inputs), such as an identity of a first resolution with an increased accuracy.
  • each respective model 118 arrives at a corresponding evaluation when provided a respective data set.
  • each respective model 118 independently arrives at a result and then the result of each respective model 118 is collectively verified through a comparison or amalgamation of the models 118. From this, a cumulative result is provided by the models 118.
  • the present disclosure is not limited thereto. DB1/ 147040794.2 45 Attorney Ref.
  • a respective model 118 is tasked with performing a corresponding activity.
  • the task performed by the respective model 118 includes, but is not limited to, evaluating a training data set (e.g., first plurality of graphical data 108-1 of Figure 2A, block 402 of Figure 4A), determining a desired threshold performance (e.g., block 428 of Figure 4C), determining if the desired threshold performance is satisfied (e.g., block 456 of Figure 4E, block 458 of Figure 4E, block 466 of Figure 4E), matching the desired threshold performance to a training resolution (e.g., block 456 of Figure 4E)), generating one or more instructions for terminating decoding of the encoded byte stream at a client device (e.g., block 458 of Figure 4E, block 466 of Figure 4E, etc.), any combination thereof.
  • a training data set e.g., first plurality of graphical data 108-1 of Figure 2A, block 402 of Figure 4A
  • each respective model 118 of the present disclosure makes use of 10 or more parameters, 100 or more parameters, 1000 or more parameters, 10,000 or more parameters, or 100,000 or more parameters. In some embodiments, each respective model 118 of the present disclosure cannot be mentally performed.
  • the codec module 122 is configured to retain a plurality of codecs (e.g., first codec 124-1, second codec 124-2, ..., codec H 124-H of Figure 2B). For instance, in some embodiments, the codec module 122 is configured to retain a plurality of codecs including between three and 20 codecs 124, between 5 and 40 codecs, between 15 and 100 codecs, or between 25 and 150 codecs.
  • the codec module 122 is configured to retain between three and 20 codecs 124, between 5 and 40 codecs, between 15 and 100 codecs, or between 25 and 150 codecs. In some embodiments, the codec module 122 is configured to retain at least 5 codecs, at least 15 codecs, at least 25 codecs, at least 50 codecs, at least 100 codecs, or at least 150 codecs. In some embodiments, the codec module 122 is configured to retain at most 5 codecs, at most 15 codecs, at most 25 codecs, at most 50 codecs, at most 100 codecs, or at most 150 codecs.
  • each codec 124 includes one or more specific instructions for converting graphical data 108 into an encoded byte stream (e.g., encoded byte stream 114 of Figure 2A) in accordance with a corresponding encoder (e.g., first encoder 126-1, second encoder 126-2, ..., encoder H 126-H of Figure 2B, etc.) and reconstructing the plurality of graphical data from the encoded byte stream in accordance with a corresponding decoder (e.g., first decoder 128-1, second decoder 126-2, ..., decoder H 128-H of Figure 2B, etc.).
  • DB1/ 147040794.2 46 Attorney Ref.
  • a respective encoder 126 includes one or more instructions to encode (e.g., progressively encode) a respective plurality of graphical data 108 into an encoded byte stream 114 through a sampling resolution of the respective plurality of graphical data.
  • the encoded byte stream 114 includes a plurality of sequence scans, which is independent of a corresponding modality and/or a corresponding resolution (e.g., sampling resolution) property 112 associated with the respective plurality of graphical data 108.
  • a first plurality of graphical data 108-1 includes M 2D images with native resolutions Xi ⁇ i ⁇ ⁇ 1,2, ... ,M ⁇ .
  • the respective encoder 126 imposes a condition that the native resolution of an initial terminal scan in the plurality of sequence scans exceeds a minimum of X (Xmin), in order to recover sufficient information from evaluation of the initial terminal scan.
  • a respective decoder 128 includes one or more instructions to decode an encoded byte stream 114 into a respective scan in a plurality of sequence scans of the encoded byte stream 114 with a corresponding native resolution.
  • the respective decoder 128 is able to decode the encoded byte stream and complete reconstruction of a respective plurality of graphical data associated with the encoded byte stream.
  • a high-throughput JPEG 2000 (“HTJ2K”) codec (e.g., first codec 124-1 of Figure 2B) is an extension to a JPEG 2000 (“J2K”) codec (e.g., second codec 124-2 of Figure 2B), that, unlike an original JPEG standard codec (e.g., third codec 124-3 of Figure 2B), enables faster decoding (e.g., using first decoder 128-1), higher quality, lossless encoding with high scalability (e.g., using first encoder 126-1), and progressive decoding of images through an encoding (e.g., compression) and decoding (e.g., decompression) of a plurality of graphical data 108.
  • J2K JPEG 2000
  • the HTJ2K codec 124 utilizes a lightweight and efficient block-coder model that yields up to a factor of 10-30 increase in throughput for lossless coding and a 6-10% increase in efficiency when compared to J2K. See Taubman et al., 2020, “High Throughput JPEG 2000 (HTJ2K) and the JPH File Format: A Primer,” available at ds.jpeg.org/whitepapers/jpeg-htj2k-whitepaper.pdf (accessed April 17, 2023), which is hereby incorporated by reference its entirety for all purposes.
  • HTJ2K lower computational complexity and support for parallelism results in a throughput on par, or even better, than the heavily optimized original JPEG standard in single and multi-threaded applications.
  • the HTJ2K codec 124 is utilized in a workflow by extending a DICOM transfer syntax standard to add support for HTJ2K. See Hafey et al., 2022, “HTJ2K Transfer Sytax Support,” available at dicomstandard.org/docs/librariesprovider2/default-documentlibrary/htj2k-transfer-syntax- support.docx (accessed April 17, 2023).
  • a client application 130 is a group of instructions that, when executed by the processor 174, generates content for presentation to the user (e.g., graphical data 108-1 of Figure 11, etc.), such as a result provided by one or more models 118.
  • the client application 130 generates content in response to one or more inputs received from the user through the graphical data system 200, such as the inputs 180 of the graphical data system 200.
  • Each of the above identified modules and applications correspond to a set of executable instructions for performing one or more functions described above and the methods described in the present disclosure (e.g., the computer-implemented methods and other information processing methods described herein).
  • modules e.g., sets of instructions
  • the memory 192 optionally stores a subset of the modules and data structures identified above.
  • the memory 192 stores additional modules and data structures not described above. DB1/ 147040794.2 48 Attorney Ref.
  • FIG. 2A and 2B is only one example of a graphical data system, and that the graphical data system 200 optionally has more or fewer components than shown, optionally combines two or more components, or optionally has a different configuration or arrangement of the components.
  • the various components shown in Figures 2A and 2B are implemented in hardware, software, firmware, or a combination thereof, including one or more signal processing and/or application specific integrated circuits.
  • FIG. 3 a description of an exemplary client device 300 that can be used with the present disclosure is provided.
  • a client device 300 includes a smart phone (e.g., an iPhone, an Android device, etc.), a laptop computer, a tablet computer, a desktop computer, a wearable device (e.g., a smart watch, a heads-up display (HUD) device, etc.), a television (e.g., a smart television), or another form of electronic device such as a gaming console, a stand-alone device, and the like.
  • a smart phone e.g., an iPhone, an Android device, etc.
  • a laptop computer e.g., a tablet computer, a desktop computer
  • a wearable device e.g., a smart watch, a heads-up display (HUD) device, etc.
  • a television e.g., a smart television
  • another form of electronic device such as a gaming console, a stand-alone device, and the like.
  • the client device 300 illustrated in Figure 3 has one or more processing units (CPU’s) 272, a network or other communications interface 274, a memory 292 (e.g., random access memory), a user interface 276, the user interface 276 including a display 278 and input 280 (e.g., keyboard, keypad, touch screen, etc.), optional audio circuitry, an optional speaker, an optional microphone, an optional input/output (I/O) subsystem, one or more communication busses 270 for interconnecting the aforementioned components, and a power system (e.g., power supply) for powering the aforementioned components.
  • the input 280 is a touch-sensitive display, such as a touch- sensitive surface.
  • the user interface 276 includes one or more soft keyboard embodiments.
  • the soft keyboard embodiments include standard (QWERTY) and or non-standard configurations of symbols on the displayed icons.
  • the input 280 and/or the user interface 276 is utilized by an end-user of the respective client device 300 (e.g., a respective subject) to input various information (e.g., a text object within a message) to the respective client device.
  • the client device 300 illustrated in Figure 3 optionally includes, in addition to accelerometer(s), a magnetometer, and a global positioning system (GPS or GLONASS or other global navigation system) receiver for obtaining information concerning a current location (e.g., a latitude, a longitude, an elevation, etc.) and/or an orientation (e.g., a portrait or a landscape orientation of the device) of the client device 300.
  • GPS global positioning system
  • GLONASS global navigation system
  • the location of the client device sets a respective codec 124 for encoding a respective graphical data 108.
  • the present disclosure is not limited thereto.
  • the client device 300 illustrated in Figure 3 is only one example of a multifunctional device that may be used for receiving an encoded byte stream 114, decoding the encoded byte stream 114, communicating a request for an evaluation of graphical data 108, terminating decoding of the encoded byte stream 114, or a combination thereof.
  • the client device 300 optionally has more or fewer components than shown, optionally combines two or more components, or optionally has a different configuration or arrangement of the components.
  • the various components shown in Figure 3 are implemented in hardware, software, firmware, or a combination thereof, including one or more signal processing and/or application specific integrated circuits.
  • client device 300 is a desktop or laptop computer.
  • the memory 292 of the client device 300 illustrated in Figure 3 optionally includes high-speed random access memory and optionally also includes non-volatile memory, such as one or more magnetic disk storage devices, flash memory devices, or other non-volatile solid- state memory devices. Access to the memory 292 by other components of the client device 300, such as CPU(s) 272 is, optionally, controlled by the memory controller. [00211] In some embodiments, the one or more CPU(s) 272 run or execute various software programs and/or sets of instructions stored in the memory 292, such as the client application 230, to perform various functions for the client device 300 and process data. [00212] In some embodiments, the CPU(s) 272 and the memory controller are implemented on a single chip.
  • the CPU(s) 272 and the memory controller are implemented on separate chips.
  • the audio circuitry, the optional speaker, and the optional microphone provide an audio interface between the respective subject and the client device 300, enabling the client device to provide a message that include audio data provided through the audio circuitry, the optional speaker, and/or the optional microphone.
  • the audio circuitry receives audio data from the peripherals interface, converts the audio data to electrical signals, and transmits the electrical signals to the speaker.
  • the speaker converts the electrical signals to human-audible sound waves.
  • the audio circuitry also receives electrical signals converted by the microphone from sound waves.
  • the audio circuitry converts the electrical DB1/ 147040794.2 50 Attorney Ref.
  • the client device 300 optionally also includes one or more optical sensors.
  • the optical sensor(s) optionally include charge-coupled device (CCD) or complementary metal-oxide semiconductor (CMOS) phototransistors.
  • CCD charge-coupled device
  • CMOS complementary metal-oxide semiconductor
  • the optical sensor(s) receive light from the environment, projected through one or more lens, and converts the light to data representing an image.
  • the optical sensor(s) optionally capture still images and or video.
  • an optical sensor is disposed on a back end portion of the client device 300 (e.g., opposite the display 278 on a front end portion of the client device 300) so that the input 280 is enabled for use as a viewfinder for still (e.g., digital image, etc.) and or video graphical data acquisition.
  • the memory 292 of the client device 300 stores: • an operating system 202 that includes procedures for handling various basic system services; • an electronic address 204 associated with the client device 300; • a model library 216 that retains a plurality of models (e.g., first model 118-1, second model 118-2, ..., model F 118-X of Figure 2B), in which each respective model 118 is configured for evaluating, at least in part, a plurality of graphical data 108 in accordance with one or more parameters of the model 118 (e.g., first parameter 120-1, second parameter 120- 2, ..., parameter D 120-D of first model 118-1 of Figure 2B); • a codec module 222 that retains a plurality of codecs (e.g., first codec 124-1, second codec 124-2, ..., codec H 124-H of Figure 2B), each codec 124 in the plurality of codecs 124 including specific instructions for converting graphical data 108 into
  • a client device 300 preferably includes an operating system 202 that includes procedures for handling various basic system services.
  • the operating system 202 e.g., iOS, ANDROID, DARWIN, RTXC, LINUX, UNIX, OS X, WINDOWS, or an embedded operating system such as VxWorks
  • the operating system 202 includes various software components and or drivers for controlling and managing general system tasks (e.g., memory management, storage device control, power management, etc.) and facilitates communication between various hardware and software components.
  • An electronic address 204 is associated with each client device 300, which is utilized to at least uniquely identify the client device from other devices and components of the system 100.
  • the electronic address 204 of the client device 300 has the same functionality as the electronic address 104 of the graphical data system 200.
  • the client device includes a model library 216 that stores a plurality of models 118 (e.g., classifiers, regressors, clustering, etc.).
  • the model library 216 of the client device 300 has the same functionality as the model library 116 of the graphical data system 200.
  • the present disclosure is not limited thereto.
  • the client device includes a codec module 222 that stores a plurality of codecs 124.
  • the codec module 222 of the client device 300 has the same functionality as the codec module 122 of the graphical data system 200.
  • the present disclosure is not limited thereto.
  • the client application 230 is a group of instructions that, when executed by the processor 272, generates content for presentation to the respective subject (e.g., graphical data set 108-2 of Figure 11), such as a result of a corresponding report generated by one or more models 118.
  • the client application 230 generates content in response to one or more inputs received from the respective subject through the client device 300, such as the inputs 280 of the client device 300.
  • the client application 230 of the client device 300 has the same functionality as the client application 130 of the graphical data system 200.
  • the present disclosure is not limited thereto.
  • the client device 300 has any or all of the circuitry, hardware components, and software components found in the system depicted in Figure 3. In the interest of brevity and clarity, only a few of the possible components of the client device 300 are shown to better emphasize the additional software modules that are installed on the client device.
  • Each of the above identified modules and applications correspond to a set of executable instructions for performing one or more functions described above and the methods described in the present disclosure.
  • These modules e.g., sets of instructions
  • the memory 292 optionally stores a subset of the modules and data structures identified above.
  • the memory 292 stores additional modules and data structures not described above.
  • Figure 3 illustrates only an example of the client device 300, and that the client device 300 optionally has more or fewer components than shown, optionally combines two or more components, or optionally has a different configuration or arrangement of the components.
  • the various components shown in Figure 3 are implemented in hardware, software, firmware, or a combination thereof, including one or more signal processing and/or application specific integrated circuits.
  • the client device 300 can be a single device that includes all the functionality of the client device 300.
  • the client device 300 can also be a combination of multiple devices.
  • the functionality of the client device 300 may be spread across any number of networked computers and/or reside on each of several networked computers and/or by hosted on one or more virtual machines and/or containers at a remote location accessible across a communications network (e.g., communications network 186, network interface 284, or both).
  • a communications network e.g., communications network 186, network interface 284, or both.
  • Various modules in a memory 192 of a graphical data system 200 e.g., graphical data store 106 of Figure 2A, model library 116 of Figure 2B, codec module 122 of Figure 2B, etc.
  • a memory 292 of a client device 300200 e.g., model library 216 of Figure 3, codec module 222 of Figure 3, etc.
  • the processes of a method of the present disclosure can be encoded in a single module or any combination of modules.
  • the method 400 optimizes encoding of graphical data (e.g., graphical data 108 of Figure 2A), communication of graphical data (e.g., encoded byte stream 114 of Figure 2A), decoding of graphical data, or a combination thereof that is communicated via a communication network (e.g., communication network 186 of Figure 1).
  • graphical data e.g., graphical data 108 of Figure 2A
  • communication of graphical data e.g., encoded byte stream 114 of Figure 2A
  • decoding of graphical data e.g., a combination thereof that is communicated via a communication network (e.g., communication network 186 of Figure 1).
  • the method 400 communicates an encoded byte stream (e.g., second encoded byte stream 114-2 of Figure 2A) in accordance with a prior training using a different graphical data set (e.g., first graphical data 108-1 of Figure 2A) associated with the encoded byte stream, a use case defined by a request for the encoded byte stream 114 (e.g., defined by a request received from a second computer system such as client device 300 of Figure 1), an aspect of a client device 300 associated with the request for the encoded byte stream 114, a required resolution for using the graphical data 108 of the encoded byte stream 114, an aspect of a model (e.g., model 118 of Figure 2B) to which the graphical data 108 is applied, or a combination thereof.
  • a model e.g., model 118 of Figure 2B
  • the method 400 allows for communicating only an optimal resolution, or corresponding scan in a plurality of sequence scans of the encoded byte stream 114 that matches a desired threshold performance. Accordingly, in some embodiments, the method 400 allows or optimizing cost (e.g., data storage costs, data capacity costs, etc.), network bandwidth, turnaround time, or a combination thereof for communicating graphical information through the encoded byte stream 114. Moreover, in some embodiments, the method 400 allows for terminating communication of the encoded byte stream in accordance with a determination that a first resolution of the encoded byte stream decoded at the second computer system satisfies a DB1/ 147040794.2 54 Attorney Ref.
  • cost e.g., data storage costs, data capacity costs, etc.
  • Block 402. Referring to block 402 of Figure 4A, in various embodiments, the method 400 is provided at a computer system (e.g., graphical data system 200 of Figures 2A and 2B, client device 300 of Figure 3, etc.).
  • the computer system 200 includes one or more processors (e.g., CPU 172 of Figure 2A, CPU 272 of Figure 3, etc.) and a memory (e.g., memory 192 of Figures 2A and 2B, memory 392 of Figure 3, etc.), such a first memory coupled to the one or more processors.
  • processors e.g., CPU 172 of Figure 2A, CPU 272 of Figure 3, etc.
  • a memory e.g., memory 192 of Figures 2A and 2B, memory 392 of Figure 3, etc.
  • the memory includes one or more programs configured to be executed by the one or more processors (e.g., graphical data store 106 of Figure 2A, model library 116 of Figure 2B, codec module 122 of Figure 2B, model library 216 of Figure 3, codec module 222 of Figure 3, etc.).
  • the method 400 requires utilization of a computer system, such as in order to evaluate the graphical data 108 and communicate the graphical data 108 via the communication network 186, and, therefore, cannot be mentally performed.
  • the method 400 allows for parallelism, which allows for encoding (e.g., using first encoder 126-1 of Figure 2B), communicating an encoded byte stream, decoding the encoded byte stream (e.g., using second decoder 126-2 of Figure 2B), or a combination thereof simultaneously using a plurality of graphical data 108 using a plurality of CPUs 172.
  • the present disclosure is not limited thereto.
  • Block 404 Referring to block 404, in some embodiments, the method 400 includes obtaining, in electronic form, a training data set.
  • the obtaining is performed by one or more models (e.g., first model 118-2, second model 118-2, model 118 F 118-F of Figure 2B, etc.), which utilize the training data as input features.
  • the one or more models 118 include at least one feature extraction model 118, such as a first segmentation model 118-1 or a second classification model 118-2.
  • the present disclosure is not limited thereto.
  • the obtaining the training data set includes accessing a respective plurality of graphical data 108 associated with the first computer system.
  • the training data set is obtained by accessing the first graphical data 108-1 stored locally at the first computer system.
  • the present disclosure is not limited thereto.
  • the training data set is obtained by accessing the first graphical data 108-1 that is stored remotely from the first computer system (e.g., stored at a first client device 300-1 of Figure 1). DB1/ 147040794.2 55 Attorney Ref. No.: 115834-5037-WO [00231]
  • the training data set is selected, at least in part, in accordance with an aspect of a respective model 118 the training data set is applied to, such as selecting a first training data set for application with a first model 118-1 and a second training data set for application with a second model 118-2.
  • the training data is selected, at least in part, in accordance with an aspect of a data type of the training data, such as such as by using the first training data set for application when an evaluation is associated with a first modality (e.g., an X-Ray modality) and the second training data set is used for application when the evaluation is associated with a second modality (e.g., an MRI modality).
  • a first modality e.g., an X-Ray modality
  • a second training data set is used for application when the evaluation is associated with a second modality (e.g., an MRI modality).
  • the training data set includes a first plurality of graphical data (e.g., first plurality of graphical data 108-1 of Figure 2A).
  • the first plurality of graphical data 108-1 is defined by a first plurality of characteristics (e.g., first characteristics 110-1 of Figure 2A) that includes one or more characteristics (e.g., first characteristic 112-1 of Figure 2A, third characteristic 112-3 of Figure 2A).
  • the one or more characteristics 112 includes a sampling resolution (e.g., first characteristic 112-1 of Figure 2A) of the first plurality of graphical data.
  • the sampling resolution includes a number of pixels or voxels that is utilized to visualize a respective graphical data 108 (e.g., visualize a respective 2D image).
  • the sampling resolution is defined by a processed utilized to capture the respective graphical data 108.
  • the first plurality of characteristics includes a first modality (e.g., second characteristic 112-2 of Figure 2A) associated with a capture of the first plurality of graphical data 108-1.
  • a first modality e.g., second characteristic 112-2 of Figure 2A
  • each respective graphical data 108 is associated with a corresponding modality (e.g., the first modality in a plurality of modalities, the first modality in one or more modalities, etc.).
  • Each respective modality is associated with a unique process and/or technique utilized for capturing the respective graphical data 108.
  • a first modality is associated with the respective graphical data 108 captured by a single-photo emission computed tomography (SPECT) process and a second modality is associated with the respective graphical data 108 captured by a magnetic resonance imaging (MRI) process.
  • SPECT single-photo emission computed tomography
  • MRI magnetic resonance imaging
  • a respective characteristic in the first plurality of characteristics describes a fine-grain aspect of the first plurality of graphical data, such as a particular aspect of a first digital image in a plurality of digital images associated with the DB1/ 147040794.2 56 Attorney Ref. No.: 115834-5037-WO first plurality of graphical data.
  • the respective characteristic in the first plurality of characteristics describes a course-grain aspect of the first plurality of graphical data, such as a particular aspect of the plurality of digital images associated with the first plurality of graphical data.
  • the first plurality of characteristics allow for tuning encoding of a respective graphical data based on either course-grain or fine-grain characteristics associated with the respective graphical data.
  • the present disclosure is not limited thereto.
  • the first plurality of graphical data 108-1 includes one or more images taken of a first biological sample.
  • a respective biological sample is a living or non-living cellular organism.
  • the respective biological sample is a living or non-living human, such as a male or female human of any stage (e.g., a man, a woman, a child, etc.).
  • the respective biological sample includes a 3D body (e.g., a human subject, a flora of a region of interest, an inanimate object, etc.).
  • the first plurality of graphical data 108-1 represents a first biological sample during a first epoch.
  • the second one or more graphical data represents a second biological sample during a second epoch.
  • the first biological sample includes a first 3D body at a first spacetime and the second biological sample includes a second 3D body, different from the first 3D body, at a second spacetime different from the first spacetime.
  • the first biological sample is a first chest X-Ray of a first human subject captured on December 1, 2000
  • the second biological sample is a second chest X-Ray of a second human captured on January 1, 2023.
  • the present disclosure is not limited thereto.
  • the first plurality of graphical data 108-1 includes 2D graphical data, 3D graphical data, 4D graphical data, or a combination thereof.
  • the first plurality of graphical data includes p- dimensional data, in which p is an integer greater than or equal to two (e.g., four, five, six, fifteen, etc.).
  • the 2D graphical data includes a corresponding plurality of pixels that collectively depict one or more digital images.
  • the 3D graphical data includes a corresponding plurality of volumetric pixels (voxels) that collectively depict one or more volumetric images.
  • graphical data 108-1 is two-dimensional (2D) (e.g., defined by x and y spatial coordinates), 3D (e.g., defined by x, y, and z spatial coordinates), four-dimensional (4D) (e.g., defined by x, y, and z spatial coordinates including a temporal coordinate), or n-dimensional (n-D).
  • 2D two-dimensional
  • 3D e.g., defined by x, y, and z spatial coordinates
  • 4D e.g., defined by x, y, and z spatial coordinates including a temporal coordinate
  • n-dimensional (n-D) e.g., defined by x, y, and z spatial coordinates including a temporal coordinate
  • n-dimensional (n-D) e.g., defined by x, y, and z spatial coordinates including a temporal coordinate
  • n-dimensional (n-D) e.g., defined by x,
  • the plurality of space-time characteristics provide a simultaneous localization and mapping (SLAM) of a trajectory.
  • the plurality of space-time characteristics is used by one or more computational models 118 to generate spatially registered base maps of a 3D body.
  • SLAM simultaneous localization and mapping
  • the plurality of space-time characteristics is used by one or more computational models 118 to generate spatially registered base maps of a 3D body.
  • the first plurality of graphical data 108-1 includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof.
  • the one or more digital images include one or more photographs, one or more computer-aided drawings (CAD), one or more maps (e.g., geographical maps, topographical maps, etc.), one or more charts, or a combination thereof.
  • CAD computer-aided drawings
  • maps e.g., geographical maps, topographical maps, etc.
  • the one or more 3D maps include one or more CT maps, one or more MRI maps, one or more PET maps, or the like, which provide 3D graphical data of an internal region (e.g., internal organs, internal tissue, internal bones) of a biological sample.
  • the one or more digital videos include one or more animations Accordingly, the method 400 allows for optimizing decoding of a variety of file types utilized to retain a respective plurality of graphical data 108 in 2D and/or 3D forms.
  • the first plurality of graphical data includes 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images.
  • the first plurality of graphical data 108-1 includes at least 50 digital images, at DB1/ 147040794.2 58 Attorney Ref. No.: 115834-5037-WO least 100 digital images, at least 1,000 digital images, at least 10,000 digital images, or at least 100,000 digital images.
  • the first plurality of graphical data 108-1 includes at most 50 digital images, at most 100 digital images, at most 1,000 digital images, at most 10,000 digital images, or at most 100,000 digital images.
  • the second one or more graphical data 108-2 includes 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images.
  • the second one or more graphical data 108-1 includes at least 50 digital images, at least 100 digital images, at least 1,000 digital images, at least 10,000 digital images, or at least 100,000 digital images.
  • the second one or more graphical data 108-1 includes at most 50 digital images, at most 100 digital images, at most 1,000 digital images, at most 10,000 digital images, or at most 100,000 digital images. [00243] Block 414.
  • the sampling resolution of the first plurality of graphical data 108-1 includes the minimum resolution associated with the first plurality of graphical data.
  • a respective model 118 imposes a condition (e.g., a first threshold performance criterion) that a native resolution of an initial terminal scan in a plurality of sequence scans must exceed the minimum resolution.
  • the minimum resolution is a predetermined resolution, such as a first predetermined resolution required for a respective graphical data to be utilized by the respective model 118 or a second predetermined resolution defined, at least in part, by an end-use of the computer system or a capture method for the graphical data.
  • the minimum resolution for 2D graphical data 108 is 8x8, 16x16, 32x32, 64x64, 128x128, or 256x256, which is required to obtain sufficient information from the first plurality of graphical data 108-1.
  • the minimum resolution for 3D graphical data 108 is 8x8x8, 16x16x16, 32x32x32, 64x64x64, 128x128x128, or 256x256x256.
  • the minimum resolution is defined by one dimension (e.g., a minimum X resolution, a minimum Y resolution, or a minimum Z resolution), by two dimensions (e.g., a minimum X resolution and a minimum Y resolution, the minimum Y resolution and a minimum Z resolution, etc.), or by three dimensions (a minimum X resolution, a minimum Y resolution, and a minimum Z resolution).
  • CT computer tomography
  • the first modality includes a computer tomography (CT) modality (e.g., an Abdomen CT modality, an angiography CT DB1/ 147040794.2 59 Attorney Ref.
  • CT computer tomography
  • an arthrography CT modality a bone CT modality, a brain CT modality, a head CT modality, a chest CT modality, a lung CT modality, a neck CT modality, a pelvis CT modality, a renal stone CT modality, a sinus CT modality, a spine CT modality, etc.
  • a digital pathology modality e.g., a magnetic resonance imaging (MRI) modality (e.g., a functional MRI modality, a breast MRI modality, a magnetic resonance angiography MRI modality, a magnetic resonance venographic MRI modality, a cardiac MRI modality, etc.), a positron emission tomography (PET) modality (e.g., a PET f-18 FDG modality, a cardiac PET modality, a PET/CT modality, etc.), a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination
  • MRI magnetic resonance imaging
  • the MRI modality includes spin-echo imaging, inversion recovering imaging, echo planar imaging, gradient echo imaging, and/or the like.
  • the sonography modality includes ultrasound imaging, doppler imaging, and/or the like.
  • a respective modality is specific to both a method of capture and a subject matter association with a region of interest (ROI) of a corresponding graphical data, such as a first abdominal CT modality, a second brain MRI, a third chest X-ray modality, a fourth cardiac abdomen ultrasound modality, a fifth fluorescence microscopy modality, a sixth retinal fundoscopy modality, and a seventh statistical graphics modality, or a combination thereof.
  • ROI region of interest
  • Block 418 Referring to block 418 of Figure 4B, in some embodiments, the method 400 includes using the first plurality of graphical data 108-1 to train a feature extraction model (e.g., first model 118-1 of Figure 2B). In some embodiments, from this training, the method 400 obtains a trained feature extraction model (e.g., second model 118-2 of Figure 2B). [00247] In some embodiments, the training the feature extraction model further obtains performance data. For instance, in some embodiments, the performance data includes a plurality of training resolutions.
  • each respective training resolution in the plurality of training resolutions is less than the sampling resolution, which allows for each respective training resolution to be associated with a reduced data size format.
  • each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model 118 in a plurality of threshold performances. For instance, in some embodiments, a first training resolution (e.g., a 32x32 resolution) is associated with a first threshold performance when using the first training resolution with a first model 118-1, and a second training resolution (e.g., a 64x32 resolution) DB1/ 147040794.2 60 Attorney Ref.
  • a first training resolution e.g., a 32x32 resolution
  • a second training resolution e.g., a 64x32 resolution
  • the feature extraction model 118 includes a segmentation model (e.g., first model 118-1 of Figure 2B) and/or a classification model (e.g., second model 118-2 of Figure 2B).
  • a segmentation model e.g., first model 118-1 of Figure 2B
  • a classification model e.g., second model 118-2 of Figure 2B
  • the one or more feature extraction models 118 includes one or more convolutional neural networks (CNN) models 118, one or more recurrent neutral networks (RNN) models 118, one or more support vector machines (SVM) models 118, one or more decision tree models 118, one or more clustering models 118, one or more random forest models 118, one or more deep belief network models 118, or a combination thereof.
  • CNN convolutional neural networks
  • RNN recurrent neutral networks
  • SVM support vector machines
  • a decision tree models 118 one or more clustering models 118
  • random forest models 118 one or more random forest models 118
  • deep belief network models 118 or a combination thereof.
  • a respective classification model 118 is utilized to assign a label to an input feature associated with a respective plurality of graphical data 108 based on a characteristic of the respective plurality of graphical data 108.
  • a respective segmentation model 118 partitions the respective plurality of graphical data 108 based on distinct features or regions present in the respective plurality of graphical data 108.
  • the feature extraction model includes a neutral network model, a support machine mode, a Na ⁇ ve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph- based model, or a combination thereof.
  • XGBoost extreme gradient boost
  • the first plurality of characteristics 110 includes one or more ground truth labels (e.g., fourth characteristic 112-4 of Figure 2A).
  • each ground truth label in the one or more ground truth labels provides an indication of a presence or an absence of a respective feature, such as an abnormality.
  • a respective ground truth label is obtained from a measurement associated with the graphical data 108, such as by a first model 118 utilized to obtain the one or more ground truth labels.
  • the respective ground truth label is obtained from a source of the respective graphical data 108, such as an administrator of a second computer system (e.g., client device 300 of Figure 1, client device 300 of Figure 3, etc.) or the first computer system (e.g., graphical data system 200 of Figure 1, graphical data system 200 of Figures 2A and 2B, etc.).
  • a source of the respective graphical data 108 such as an administrator of a second computer system (e.g., client device 300 of Figure 1, client device 300 of Figure 3, etc.) or the first computer system (e.g., graphical data system 200 of Figure 1, graphical data system 200 of Figures 2A and 2B, etc.).
  • each respective ground truth label provides an DB1/ 147040794.2 61 Attorney Ref. No.: 115834-5037-WO indication of a known corresponding outcome, output, or conclusion based on an associated input of the respective graphical data 108.
  • the first plurality of graphical data 108-1 is used to train the feature extraction model 118 against the one or more ground truth labels (e.g., first characteristic 112-1 of Figure 2A).
  • the method 400 includes receiving a request to evaluate a second one or more graphical data (e.g., second graphical data 108-2 of Figure 2A).
  • the request to evaluate the second one or more graphical data 108 is a request to perform the evaluation at a second computer system.
  • the method 400 allows for optimizing decoding of the second one or more graphical data 108-2 at the second computer system based on using the first plurality of graphical data 108 with the feature extraction model 118 at the first computer system.
  • the request to evaluate the second one or more graphical data 108 is a request to perform the evaluation at the first computer system.
  • the second one or more graphical data 108-2 is defined by a second plurality of characteristics (e.g., second characteristics 110-2 of Figure 2A.
  • the second plurality of characteristics is different than the first plurality of characteristics of the first plurality of graphical data 108-1.
  • the present disclosure is not limited thereto.
  • the second plurality of characteristics 110-2 includes one or more characteristics including the first modality associated with a capture of the second one or more graphical data 108-1. Accordingly, in some embodiments, the process of capturing the first plurality of graphical data 108-1 and the second one or more graphical data 108-2 are the same (e.g., both the first plurality of graphical data 108-1 and the second one or more graphical data 108-2 were captured by an MRI process, both the first plurality of graphical data 108-1 and the second one or more graphical data 108-2 were captured by CT process, etc.). [00254] In some embodiments, the second plurality of characteristics 110-2 includes one or more characteristics including a desired threshold performance.
  • the desired performance includes a threshold accuracy and/or a threshold precision that must be satisfied by a respective training resolution when applying the second one or more graphical data 108-2 to a respective feature extraction model 118.
  • the desired DB1/ 147040794.2 62 Attorney Ref. No.: 115834-5037-WO threshold performance is based on a characteristic of the second one or more graphical data 108-2, the first plurality of graphical data 108-1, a parameter 120 of the respective feature extraction model 118, or a combination thereof.
  • the present disclosure is not limited thereto.
  • the desired threshold performance is based on a first evaluation of a first training resolution.
  • the desired threshold performance is based, at least in part, on using the sampling resolution with a respective feature extraction model 118 in one or more feature extraction models at the first computer system.
  • the desired threshold performance defines a quality metric that is a minimum ability to extract information from any resolution (e.g., any scan in a plurality of sequence scans) from an encrypted byte stream 114 that maintains full information efficiency.
  • the second one or more graphical data 108-2 includes one or more images taken of a second biological sample different than the first biological sample. As such, in some such embodiments, different biological samples are utilized to create a respective graphical data 108.
  • the first biological sample is a first human subject and the second biological sample is a second human. Accordingly, the method 400 allows for training a model 118 using a first biological sample and evaluating the second biological sample based on the training with the first biological sample.
  • the first biological sample is an initial sample used to validate and/or train a respective model 118, such as in order to identify one or more patterns associated with the first biological sample
  • the second biological sample is a real-world dataset.
  • a human 3D body with each sub-body in one or more bodies defined at a point of articulation of the human 3D body.
  • the systems and methods of the present disclosure obtain a first plurality of graphical data 108-1 associated with a first sub-body in the one or more sub-bodies and a second one or more graphical data 108-2 associated with a second sub-body in the one or more sub-bodies (e.g., first graphical data is of a face, second digital data is of a foot, etc.).
  • a third plurality of graphical data 108-3 is generated based on the first and second pluralities of graphical data 108, in which the third plurality of graphical is associated with a combination of the first and second pluralities of graphical 108.
  • the present disclosure is not limited thereto. DB1/ 147040794.2 63 Attorney Ref.
  • the second one or more graphical data 108-2 includes 3D graphical data.
  • each scan in the plurality of sequence scans of the 3D graphical data includes a respective 2D layer of the 3D graphical data.
  • each 2D layer is formed by applying a splicing technique to the 3D graphical data.
  • a first layer includes 1 x 10 2 points
  • a second layer includes 1 x 10 3 points
  • a third layer includes 1 x 10 5 points, and the like.
  • the present disclosure is not limited thereto.
  • Block 434 the second one or more graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof.
  • the method 400 includes point cloud pre-processing, spatial clustering segmentation based on K-dimensional tree, acquisition of structural parameters, calculation of volume based on multiple regression analysis, and the like.
  • Block 436 Referring to block 436, in some embodiments, the second plurality of characteristics further includes a respective capacity of the communication network 186.
  • a first epoch is required to communicate a respective plurality of graphical data 108 via a communication network 186, such that the respective capacity of the communication network 186 limits an ability to communicate a respective plurality of graphical data 108 through an encoded byte stream.
  • the respective capacity of the communication network 186 includes a bandwidth of the communication network 186, which defines an amount of data that is communicated per unit time (e.g., 100 megabyte (Mb) per second, etc.).
  • the bandwidth of the communication network 186 changes with time.
  • the desired threshold performance dynamically modifies to similar to the change in the bandwidth of the communication network 186.
  • the respective capacity of the communication network 186 includes a latency of the communication network 186, which defines an epoch required to communicate a respective data packet (e.g., a respective scan in a plurality of sequence scans of the encoded byte stream).
  • the second plurality of characteristics further includes a respective capacity of the first computer system (e.g., graphical data system 200 of Figures 2A and 2B, client device 300-R of Figure 1, client device 300 of Figure 3, etc.).
  • the respective capacity of the first computer system is a hardware limitation associated with utilizing a respective model 118 at the first computer system.
  • the respective capacity of the first computer system includes a processing power of the first computer system, a memory requirement of first computer system, a storage requirement of first computer system, a heat dissipation requirement of the first computer system, a power consumption requirement of first computer system, or a combination thereof.
  • the present disclosure is not limited thereto.
  • the respective graphical data includes one or more images of the biological sample on the substrate is obtained. Each of the one or more images includes a corresponding plurality of pixels in the form of an array of pixel values.
  • the array of pixel values includes at least a least 100, 10,000, 100,000, 1 x 10 6 , 2 x 10 6 , 3 x 10 6 , 5 x 10 6 , 8 x 10 6 , 10 x 10 6 , or 15 x 10 6 pixel values.
  • the single intensity threshold is determined using Otsu’s method, where the first heuristic classifier identifies a threshold that minimizes intra-class variance or equivalently maximizes inter-class variance.
  • Otsu’s method uses a discriminative analysis that determines an intensity threshold such that binned subsets of pixels in the plurality of pixels are as clearly separated as possible.
  • Each respective pixel in the plurality of pixels is binned or grouped into different classes depending on whether the respective intensity value of the respective pixel falls over or under the intensity threshold.
  • bins are represented as a histogram, and the intensity threshold is identified such that the histogram can be assumed to have a bimodal distribution (e.g., two peaks) and a clear distinction between peaks (e.g., valley).
  • the plurality of pixels in the obtained image is filtered such that pixels including a pixel intensity above the intensity threshold are considered to be foreground and are converted to white (e.g., uint8 value of 1), while pixels including a pixel DB1/ 147040794.2 65 Attorney Ref.
  • No.: 115834-5037-WO intensity below the intensity threshold are considered to be background and are converted to black (e.g., uint8 value of 0).
  • an outcome of a heuristic classifier using Otsu’s method is illustrated (e.g., using display 182 of Figure 2A), which depicts a thresholded image (e.g., a mask or a layer) after conversion of the acquired image, where each pixel in the plurality of pixels is represented as either a white or a black pixel.
  • Otsu’s method is an example of a binarization method using global thresholding.
  • Otsu’s method is robust when the variances of the two classes (e.g., foreground and background) are smaller than the mean variance over the obtained image as a whole.
  • the first heuristic classifier is a binarization method other than Otsu’s method.
  • the first heuristic classifier is a global thresholding method other than Otsu’s method or an optimization-based binarization method.
  • a global thresholding method is performed by determining the intensity threshold value manually (e.g., via default or user input). For example, an intensity threshold can be determined at the middle value of the grey-scale range (e.g., 128 between 0- 255).
  • the intensity threshold value is determined automatically using a histogram of grey-scale pixel values (e.g., using the mode method and/or P-tile method).
  • a histogram of grey-scale pixel values can include a plurality of bins (e.g., up to 256 bins for each possible grey-scale pixel value 0- 255), and each respective bin is populated with each respective pixel having the respective grey-scale pixel value.
  • the plurality of bins has a bimodal distribution and the intensity threshold value is the grey-scale pixel value at which the histogram reaches a minimum (e.g., at the bottom of the valley).
  • an intensity threshold value is determined by estimating the level of background noise (e.g., in imaging devices including but not limited to fluorescence microscopy). Background noise can be determined using control samples and/or unstained samples during normalization and pre-processing. DB1/ 147040794.2 66 Attorney Ref.
  • the assignment of a respective pixel to one of two classes is determined by calculating the relative closeness of the converted pixel value to the original pixel value, as well as the relative closeness of the converted pixel value of the respective pixel to the converted pixel values of neighboring pixels (e.g., using a Markov random field).
  • Optimization-based methods thus include a smoothing filter that reduces the appearance of small punctate regions of black and/or white and ensures that local neighborhoods exhibit relatively congruent results after binarization. See, Uchida, 2013, “Image processing and recognition for biological images,” Develop.
  • the first or second heuristic classifier include a smoothing method to minimize or reduce noise between respective pixels in a local neighborhood by filtering for differences in pixel intensity values.
  • smoothing is performed in a plurality of pixels in grey-scale space.
  • applicable smoothing methods include, but are not limited to, blurring filters, median filters, and/or bilateral filters.
  • a blurring filter minimizes differences within a local neighborhood by replacing the pixel intensity values at each respective pixel with the average intensity values of the local neighborhood around the respective pixel.
  • a median filter utilizes a similar method, but replaces the pixel intensity values at each respective pixel with the median pixel values of the local neighborhood around the respective pixel.
  • blurring filters and median filters cause image masks to exhibit “fuzzy” edges
  • a bilateral filter preserves edges by determining the difference in intensity between pixels in a local neighborhood and reducing the smoothing effect in regions where a large difference is observed (e.g., at an edge). See, Uchida, 2013, “Image processing and recognition for biological images,” Develop. Growth Differ.55, 523-549, doi:10.1111/dgd.12054, which is hereby incorporated herein by reference in its entirety.
  • a global thresholding method is further applied to an image mask including the outcome of a local intensity gradient filter represented as an array (e.g., a matrix) of grey-scale pixel values.
  • the local intensity gradient array is binarized into two classes using Otsu’s method, such that each pixel in the plurality of pixels is converted to a white or a black pixel (e.g., having pixel value of 1 or 0, respectively), representing foreground or background, respectively.
  • Otsu’s method such that each pixel in the plurality of pixels is converted to a white or a black pixel (e.g., having pixel value of 1 or 0, respectively), representing foreground or background, respectively.
  • DB1/ 147040794.2 67 Attorney Ref.
  • a second heuristic classifier is a two-dimensional Otsu’s method, which, in some instances, provides better image segmentation for images with high background noise.
  • the grey-scale intensity value of a respective pixel is compared with the average intensity of a local neighborhood.
  • an average intensity value is calculated for a local neighborhood within a fixed distance radius around the respective pixel, and each pair of intensity values (e.g., a value averaged over the local neighborhood and a value for the respective pixel) are binned into a discrete number of bins.
  • the local neighborhood is defined by a disk including a radius of fixed length (e.g., 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 pixels, between 10 and 50 pixels, between 50 and 100 pixels, between 100 and 200 pixels, or more than 200 pixels).
  • a radius of fixed length e.g., 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 pixels, between 10 and 50 pixels, between 50 and 100 pixels, between 100 and 200 pixels, or more than 200 pixels.
  • the graph cut segmentation algorithm attempts to compute the alpha values for T U given input regions for T B and T F , by creating an alpha-matte that reflects the proportion of foreground and background for each respective pixel in a plurality of pixels as an alpha value between 0 and 1, where 0 indicates background and 1 indicates foreground.
  • an alpha value is computed by transforming a grey-scale pixel value (e.g., for an 8-bit single-channel pixel value between 0 DB1/ 147040794.2 68 Attorney Ref. No.: 115834-5037-WO and 255, the pixel value is divided by 255).
  • Graph cut is an optimization-based binarization technique as described above, which uses polynomial-order computations to achieve robust segmentation even when foreground and background pixel intensities are poorly segregated.
  • Rother et al. 2004, “‘GrabCut’ – Interactive Foreground Extraction using Iterated Graph Cuts,” ACM Transactions on Graphics.23(3):309-314, doi:10.1145/1186562.1015720, which is hereby incorporated herein by reference in its entirety.
  • Boykov and Jolly 2001, “Interactive graph cuts for optimal boundary and region segmentation of objects in N-D images,” Proc. IEEE Int. Conf.
  • the trimap is user specified.
  • the trimap is initialized using the plurality of heuristic classifiers as an initial tissue detection function.
  • the set of classes is provided to the graph cut segmentation algorithm using an alternate trimap that is a combination or substitution of the above implementations that will be apparent to one skilled in the art.
  • the segmentation algorithm is a GrabCut segmentation algorithm.
  • the GrabCut segmentation algorithm is based on a graph cut segmentation algorithm, but includes an iterative estimation and incomplete labelling function that limits the level of user input required and utilizes an alpha computation method used for border matting to reduce visible artefacts. Furthermore, GrabCut uses a soft segmentation approach rather than a hard segmentation approach. Unlike graph cut segmentation algorithms, GrabCut uses Gaussian Mixture Models (GMMs) instead of histograms of labelled trimap DB1/ 147040794.2 69 Attorney Ref. No.: 115834-5037-WO pixels, where a GMM for a background and a GMM for a foreground are full-covariance Gaussian mixtures with K components.
  • GMMs Gaussian Mixture Models
  • a unique GMM component is assigned to each pixel in the plurality of pixels from either the background or the foreground model (e.g., 0 or 1). See, Rother et al., 2004, “‘GrabCut’ – Interactive Foreground Extraction using Iterated Graph Cuts,” ACM Transactions on Graphics.23(3):309-314, doi:10.1145/1186562.1015720, which is hereby incorporated herein by reference in its entirety.
  • the GrabCut segmentation algorithm can operate either on a multi-spectral, multi-channel image (e.g., a 3-channel image) or on a single-channel image.
  • a grey-scale image is provided to the segmentation algorithm.
  • a grey-scale image is first converted to a multi-spectral, multi-channel image (e.g., RGB, HSV, CMYK) prior to input into the segmentation algorithm.
  • a multi-spectral, multi-channel color image is applied directly to the segmentation algorithm.
  • the GrabCut segmentation algorithm is applied to the image as a convolution method, such that local neighborhoods are first assigned to a classification (e.g., foreground or background) and assignations are then applied to a larger area.
  • an image including a plurality of pixels is provided to the GrabCut algorithm as a color image, using the initialization labels obtained from the plurality of heuristic classifiers, and the binary classification output of the GrabCut algorithm is used for downstream spatial analysis (e.g., on barcoded capture spots).
  • the plurality of pixels assigned with a greater probability of tissue or background is used to generate a separate construct (e.g., a matrix, array, list or vector) indicating the positions of tissue and the positions of background in the plurality of pixels.
  • the GrabCut segmentation algorithm performs binary identification of tissue and background, which is evident from the clear isolation of the tissue section overlay from the background regions.
  • the aggregated score and intensity of each respective pixel in the plurality of pixels is applied to a segmentation algorithm other than a graph cut segmentation algorithm or a GrabCut segmentation algorithm, including but not limited to, Magic Wand, Intelligent Scissors, Bayes Matting, Knockout 2, level sets, binarization, background subtraction, watershed method, region growing, clustering, active contour model (e.g., SNAKES), template matching and recognition-based method, Markov random field.
  • No.: 115834-5037-WO plurality of pixels is applied to a feature extraction algorithm (e.g., intuition and/or heuristics, gradient analysis, frequency analysis, histogram analysis, linear projection to a trained low- dimensional subspace, structural representation, and/or comparison with another image).
  • a feature extraction algorithm e.g., intuition and/or heuristics, gradient analysis, frequency analysis, histogram analysis, linear projection to a trained low- dimensional subspace, structural representation, and/or comparison with another image.
  • the aggregated score and intensity of each respective pixel in the plurality of pixels is applied to a pattern classification method including but not limited to nearest neighbor classifiers, discriminant function methods (e.g., Bayesian classifier, linear classifier, piecewise linear classifier, quadratic classifier, support vector machine, multilayer perception/neural network, voting), and/or classifier ensemble methods (e.g., boosting, decision tree/random forest).
  • discriminant function methods e.g., Bayesian
  • the plurality of pixels is in, or is converted to, grey-scale space by obtaining the image in grey-scale (e.g., a single-channel image), or by obtaining the image in color (e.g., a multi-channel image) and converting the image to grey-scale after the obtaining and prior to the running of the heuristic classifiers.
  • each respective pixel in the plurality of pixels in grey-scale space has an integer value between 0 and 255 (e.g., 8-bit unsigned integer value or “uint8”).
  • the integer value for each respective pixel in the plurality of pixels of the image in grey-scale space is transformed using e.g., addition, subtraction, multiplication, or division by a value N, where N is any real number.
  • N is any real number.
  • each respective pixel in the plurality of pixels in grey-scale space has an integer value between 0 and 255, and each integer value for each respective pixel is divided by 255, thus providing integer values between 0 and 1.
  • the plurality of pixels of the image is in grey-scale space and is transformed using contrast enhancement or tone curve alignment.
  • the running of the plurality of heuristic classifiers on the plurality of pixels includes rotating, transforming, resizing, or cropping the obtained image in grey-scale space.
  • the desired threshold performance includes a threshold inter-image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation.
  • the threshold inter- image correlation includes a first requirement for a degree of similarity between a first graphical data and a second graphical data (e.g., a first degree of similarity between a first DB1/ 147040794.2 71 Attorney Ref.
  • the threshold intra-image correlation includes a second requirement for a degree of similarity within a respective graphical data.
  • the degree of similarity is between two or more regions of a respective 2D or 3D image and/or between two or more features of the respective 2D or 3D image.
  • a relatively larger degree of similarity means that the two or more regions of the respective 2D or 3D image and/or the two or more features of the respective 2D or 3D image have a high similarity.
  • the threshold inter-and-intra- image correlation includes a requirement for a degree of similarity between two or more regions and/or two or more features of a first 2D or 3D image and of between the first 2D or 3D image and a second 2D or 3D image, which allows for the method 400 to identify patterns both within a respective 3D or 3D image or between 2D and 3D images of graphical data 108.
  • the method 400 further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data 108-2. From this encoding, the method 400 forms an encoded byte stream (e.g., first encoded byte stream 114-1 of Figure 2A).
  • the encoded byte stream includes a plurality of sequence scans through the sampling resolution, which allows an end-user to view, at a highest resolution, the sampling resolution of the second one or more graphical data 108-2.
  • each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order.
  • each rank order defines a sequence of respective sequence scans in the plurality of sequence scans.
  • the rank order defines a structured organized set that assigns each unique rank in the rank order to a position in the rank order.
  • the rank order is based on a value of a corresponding resolution.
  • each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. DB1/ 147040794.2 72 Attorney Ref.
  • the first codec 124-1 includes a predictive codec 124, an embedded codec 124, a sub-band codec 124, a block- based codec 124, a layered codec 124, a lossless codec 124, a lossy codec 124, or a combination thereof.
  • the predictive codec 124 includes one or more instructions that predict content (e.g., subject matter) for a portion of a respective plurality of graphical data 108 based on information learned from the respective plurality of graphical data 108.
  • the predictive codec the predictive codec 124 includes one or more instructions that predict a feature of second 2D image in a second one or more graphical data 108-2 based on a first 2D image in a first plurality of graphical data 108-1.
  • the present disclosure is not limited thereto.
  • the embedded codec 124 includes one or more instructions that is integrated (e.g., embedded, included) within a corresponding storage format (e.g., a file type, a container, a mountable image, etc.).
  • the sub-band codec 124 includes one or more instructions for segmenting a respective plurality of graphical data 108 into two or more bands (e.g., two or more discrete frequency bands), in which each respective band in the two or more bands is processed individually by a corresponding encoder 126 and/or a corresponding decoder 128 of the sub-band codec 124.
  • two or more bands e.g., two or more discrete frequency bands
  • the block-based codec 124, the sub-band codec 124 includes one or more instructions for segmenting a respective plurality of 2D graphical data 108, particularly video 2D graphical data, into two or more sub-blocks), in which each respective band in the two or more bands is processed individually by a corresponding encoder 126 and/or a corresponding decoder 128 of the block-based codec 124.
  • the layered codec 124 includes one or more instructions for layering a respective plurality of graphical 108 into two or more discrete layers that are collectively stored in a corresponding storage format (e.g., a file type, a container, a mountable image, etc.).
  • the lossless codec 124 utilizing a corresponding encoder 126 and a corresponding decoder 128 that communicates a respective plurality of graphical data 108 through an encoded byte stream without omitting a portion of information, which allows for maintaining high quality information within the encoded byte stream.
  • the lossy codec 124 utilizing a corresponding encoder 126 and a corresponding decoder 128 that communicates a respective plurality of graphical data 108 through an encoded byte stream by omitting a portion of information, such as a first portion of information that is not requirement to satisfy the threshold performance criterion, which allows for a reduced file size of the encoded byte stream.
  • Block 446 Referring to block 446, in some embodiments, the first codec 124-1 is selected by the first computer system (e.g., graphical data system 200 of Figure 2A and 2B, client device 300 of Figure 3, etc.).
  • the first computer system requires using the second one or more graphical data 108-2 in a first file type compatible with the first codec 124-1.
  • a first model 118-1 at the first computer system requires receiving a respective plurality of graphical data 108 in accordance with the first codec 124-1.
  • the present disclosure is not limited thereto.
  • Block 448 in some embodiments, the plurality of sequence scans includes between 2 scans and 100 scans, between 3 and 70 scans, between 4 and 50 scans, between 5 and 30 scans, between 6 and 20 scans, between 7 and 15 scans, or between 8 and 12 scans.
  • the plurality of sequenced scans includes at least 3 scans, at least 4 scans, at least 5 scans, at least 6 scans, at least 7 scans, at least 8 scans, at least 9 scans, at least 10 scans, at least 12 scans, at least 15 scans, at least 20 scans, at least 30 scans, at least 50 scans, or at least 100 scans.
  • the plurality of sequenced scans includes at most 3 scans, at most 4 scans, at most 5 scans, at most 6 scans, at most 7 scans, at most 8 scans, at most 9 scans, at most 10 scans, at most 12 scans, at most 15 scans, at most 20 scans, at most 30 scans, at most 50 scans, or at most 100 scans.
  • Block 450 Referring to block 450, in some embodiments, the plurality of sequence scans includes N scans. In some embodiments, ⁇ is ⁇ ⁇ ⁇ ⁇ 2 ⁇ ⁇ ⁇ ⁇ ⁇ 2 . In some embodiments, M is a number of digital assets associated with the second one or more graphical data.
  • ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ 1,2, .. , ⁇ is a native resolution associated with the second one or more graphical data.
  • ⁇ ⁇ ⁇ ⁇ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans.
  • the plurality of sequence scans includes N scans.
  • ⁇ is ⁇ ⁇ 2 ⁇ 1.
  • M is a number of digital assets associated with the second one or more graphical data.
  • ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ 1,2, .. , ⁇ is a native resolution associated with the second one or more graphical data.
  • ⁇ ⁇ ⁇ ⁇ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. Accordingly, in some embodiments, the method 400 allows for extracting information from any power of 2-related resolution from the encoded byte stream while maintaining the information (e.g., lossless information, about lossless information, lossy information) from the second one or more graphical data 108-2.
  • the plurality of sequence scans 1010 includes N scans, in which ⁇ is determined for each graphical data in the one or more graphical data 108 separately prior to forming the encoded byte stream map 600 and/or encoded byte stream 114, which ensured that the corresponding resolution of a sequence scan 1010 remained consistent across medical images with different resolutions as follows ⁇ where ⁇ and ⁇ are a horizontal resolution and a vertical resolution, respectively, of the graphical data 108, and ⁇ is a lower bound for the resolution of the initial terminal scan.
  • Block 454 referring to block 454, in some embodiments, the second plurality of characteristics includes a sampling resolution of the second one or more graphical data.
  • the plurality of sequence scans includes N scans.
  • N is based, at least in part, on the sampling resolution of the second one or more graphical data. For instance, in some such embodiments, N is selected based on the sampling resolution of the second one or more graphical data.
  • the second one or more graphical data 108 is progressively encoded using a corresponding encoder 126 of a lossless HTJ2K codec in the form of a plurality of sequence scans that includes N scans.
  • a respective scan is also known as a respective decomposition or a respective subset of the encoded byte stream.
  • each respective scan has a native resolution as a factor of 2 N of the full graphical data resolution (e.g., resolution associated with a final terminal scan in the plurality of sequence scans, the maximum resolution, etc.).
  • the native resolution of a respective scan in the plurality of sequence scans DB1/ 147040794.2 75 Attorney Ref. No.: 115834-5037-WO refers to a proportion of pixels of the final terminal scan in the plurality of sequence scans communication prior to any interpolation.
  • a HTJ2K encoded byte stream is communicated in the form of the plurality of sequence scans including N scans, as illustrated in Figure 6.
  • a second one or more graphical data associated with a first 2D chest x-ray (CXR) modality with a resolution of 256x256 is progressively encoded using a corresponding encoder 126 into an encoded byte stream including 3 scans in a plurality of sequence scans with native resolutions 64x64, 128x128, and 256x256, respectively.
  • the method 400 allows the HTJ2K codec 124 to not just achieve a high compression ratio, but also enables quick access to smaller resolution scans in the plurality of sequence scans of the second one or more graphical data, thus minimizing data duplication.
  • a visualization of the second one or more graphical data is formed by using about 25% of the encoded byte stream using the JPEG codec and the HTJ2K codec, respectively.
  • Block 456 the method 400 includes matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data. For instance, in some embodiments, referring briefly to chart 1200 of Figure 11, the method matches the desired threshold performance to the first training resolution in accordance with one or more logical operation based on some Boolean operation.
  • a first logical operation describes an “AND” Boolean operation that requires both elements of the first logical operation to be satisfied for a respective threshold to be deemed satisfied.
  • a second logical operation describes an “OR” Boolean operation that requires any one element of the second logical operation to be satisfied for a respective threshold to be deemed satisfied.
  • a third logical operation describes an “EXCLUSIVE OR (XOR)” Boolean operation that requires any one element of the third logical operation to be satisfied and no other element satisfied for a respective threshold to be deemed satisfied.
  • a fourth logical operation describes a singular “NOT” Boolean operation that requires absence of an element of the fourth logical operation to be satisfied for a respective threshold to be deemed satisfied.
  • a fifth logical operation describes a plural “NOT” Boolean operation that requires both absence of a first element and presence of a second element of the fifth logical operation to be satisfied for a respective threshold to be deemed satisfied.
  • a logical operation of a respective model includes a combination of one or more of the above described DB1/ 147040794.2 76 Attorney Ref. No.: 115834-5037-WO logical operations.
  • a respective logical operation includes one or more AND, OR, XOR, or NOT operations within the respective logical operation that is utilized to control a subset of a plurality of microfluid devices 300 associated with the device 100.
  • the present disclosure is not limited thereto.
  • the method matches the desired threshold performance to the first training resolution in accordance with a determination the desired threshold performance is satisfied or not satisfied by taking a cosine similarity measure or dot product of the desired threshold performance against each training resolution in the plurality of training resolutions.
  • Block 458 in some embodiments, the method 400 includes communicating, via a communication network 186, in accordance with the first codec 124, to a second computer system (e.g., client device 300 of Figure 3), the encoded byte stream 114. In some embodiments, the method 400 further includes generating and/or communicating one or more instructions to terminate decoding of the encoded byte stream 114 at the second computer system.
  • the one or more instructions instruct to terminate decoding when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution. From this communicating of the encoded byte stream 114 and the one or more instructions, the method 400 optimizing decoding of graphical data 108, such as by preventing the second computer system 200 from having to receiving excess scans that exceed the first training resolution.
  • the present disclosure is not limited thereto.
  • Block 460 Referring to block 460, in some embodiments, the first sequence scan includes less than 5% of the encoded byte stream.
  • the first sequence scan includes between 0.1% and 2% of the encoded byte stream, between 0.3% and 1.8%, between 0.5% and 1.6%, between 0.7% and 1.4%, and between 0.9% and 1.2%.
  • the first sequence scan includes at least 0.1% of the encoded byte stream, at least 0.3% of the encoded byte stream, at least 0.5% of the encoded byte stream, at least 0.7% of the encoded byte stream, at least 0.9% of the encoded byte stream, at least 1.1% of the encoded byte stream, at least 1.3% of the encoded byte stream, at least 1.5% of the encoded byte stream, at least 1.7% of the encoded byte stream, or at least 1.9% of the encoded byte stream.
  • the first sequence scan includes at most 0.1% of the encoded byte stream, at most 0.3% of the encoded byte stream, at most 0.5% of the encoded byte DB1/ 147040794.2 77 Attorney Ref. No.: 115834-5037-WO stream, at most 0.7% of the encoded byte stream, at most 0.9% of the encoded byte stream, at most 1.1% of the encoded byte stream, at most 1.3% of the encoded byte stream, at most 1.5% of the encoded byte stream, at most 1.7% of the encoded byte stream, or at most 1.9% of the encoded byte stream.
  • Block 462 referring to block 462, in some embodiments, the first sequence scan includes less than 5% of the final terminal scan.
  • the first sequence scan includes between 0.5% and 4% of the final terminal scan, between 1% and 2.5%, between 1.5% and 1.6%, between 0.7% and 1.4%, and between 0.9% and 1.2%. [00306] In some embodiment, the first sequence scan includes at least 0.1% of the final terminal scan, at least 0.3% of the final terminal scan, at least 0.5% of the final terminal scan, at least 0.7% of the final terminal scan, at least 0.9% of the final terminal scan, at least 1.1% of the final terminal scan, at least 1.3% of the final terminal scan, at least 1.5% of the final terminal scan, at least 1.7% of the final terminal scan, or at least 1.9% of the final terminal scan.
  • the first sequence scan includes at most 0.1% of the final terminal scan, at most 0.3% of the final terminal scan, at most 0.5% of the final terminal scan, at most 0.7% of the final terminal scan, at most 0.9% of the final terminal scan, at most 1.1% of the final terminal scan, at most 1.3% of the final terminal scan, at most 1.5% of the final terminal scan, at most 1.7% of the final terminal scan, or at most 1.9% of the final terminal scan.
  • Block 464 Referring to block 464, in some embodiments, the encoded byte stream 114 is associated with a first file size of the final terminal scan in the plurality of sequence scans.
  • the method 400 includes terminating the communicating of the encoded byte stream 114. For instance, in some embodiments, the method 400 terminates communicating the encoded byte stream 114 via the communication network 186 in accordance with a determination that decoding of the encoded byte stream 114 at the second computer system satisfies a first sequence scan in the plurality of sequence scans.
  • the method 400 terminates communicating the encoded byte stream 114 via the communication network 186 in accordance with a determination that decoding of the encoded byte stream 114 at the second DB1/ 147040794.2 78 Attorney Ref. No.: 115834-5037-WO computer system the first sequence scan matches or exceeds the first training resolution. In some embodiments, this determination is made at the first computer system. In some embodiments, this determination is made at the second computer system. In some embodiments, this determination is made prior to initiation of the communicating of the encoded byte stream. However, the present disclosure is not limited thereto.
  • the method 400 obtains a training data set including the first plurality of graphical data 108-1 in order to encode an encoded byte stream 114 of second one or more graphical data 108-2 in accordance with a request for an evaluation of the second one or more graphical data. From this, the method 400 allows for an optimal subset of the encoded byte stream 114, which is associated with the first training resolution, to be communicated via the communication network 186 without negatively impacting performance of a respective model 118 to which the second one or more graphical data 108-2 is applied.
  • the first sequence scan is identified in accordance with a requirement to satisfy a desired threshold performance so there is no statistically significant difference in the model performance between the final terminal scan and the first sequence scan in the plurality of sequence scans.
  • Various modules in a memory 192 of a graphical data system 200 e.g., graphical data store 106 of Figure 2A, model library 116 of Figure 2B, codec module 122 of Figure 2B, etc.
  • a memory 292 of a client device 300200 e.g., model library 216 of Figure 3, codec module 222 of Figure 3, etc.
  • the processes of a method of the present disclosure can be encoded in a single module or any combination of modules.
  • FIG. 5A through 5E there is depicted a flowchart illustrating an exemplary method 500 in accordance with some embodiments of the present disclosure.
  • the preferred parts of the method are shown in solid line boxes, whereas additional, optional, or alterative parts of the method are shown in dashed line boxes.
  • the method 500 optimizes encoding of graphical data (e.g., graphical data 108 of Figure 2A), communication of graphical data (e.g., encoded byte stream DB1/ 147040794.2 79 Attorney Ref.
  • the method 500 communicates an encoded byte stream (e.g., second encoded byte stream 114-2 of Figure 2A) in accordance with a prior training using a different graphical data set (e.g., first graphical data 108-1 of Figure 2A), a use case defined by a request for the encoded byte stream 114 (e.g., defined by a request received from a second computer system such as client device 300 of Figure 3), an aspect of a client device 300 associated with the request for the encoded byte stream 114, a required resolution for using the graphical data 108 of the encoded byte stream 114, an aspect of a model (e.g., model 118 of Figure 2B) to which the graphical data 108 is applied, or a combination thereof.
  • a model e.g., model 118 of Figure 2B
  • the method 500 allows for communicating only an optimal resolution, or corresponding scan in a plurality of sequence scans of the encoded byte stream 114 that matches a desired threshold performance. Accordingly, in some embodiments, the method 500 allows or optimizing cost (e.g., data storage costs, data capacity costs, etc.), network bandwidth, turnaround time, or a combination thereof for communicating graphical information through the encoded byte stream 114. Moreover, in some embodiments, the method 500 allows for terminating communication of the encoded byte stream in accordance with a determination that a first resolution of the encoded byte stream decoded at the second computer system satisfies a threshold resolution, such as by matching or exceeding a first resolution.
  • a threshold resolution such as by matching or exceeding a first resolution.
  • the present disclosure is not limited thereto.
  • the method 500 is provided at a computer system (e.g., graphical data system 200 of Figures 2A and 2B, client device 300 of Figure 3, etc.).
  • the computer system 200 includes one or more processors (e.g., CPU 172 of Figure 2A, CPU 272 of Figure 3, etc.) and a memory (e.g., memory 192 of Figures 2A and 2B, memory 392 of Figure 3, etc.), such a first memory coupled to the one or more processors.
  • processors e.g., CPU 172 of Figure 2A, CPU 272 of Figure 3, etc.
  • a memory e.g., memory 192 of Figures 2A and 2B, memory 392 of Figure 3, etc.
  • the memory includes one or more programs configured to be executed by the one or more processors (e.g., graphical data store 106 of Figure 2A, model library 116 of Figure 2B, codec module 122 of Figure 2B, model library 216 of Figure 3, codec module 222 of Figure 3, etc.).
  • the method 500 requires utilization of a computer system, such as in order to evaluate the graphical data 108 and communicate the graphical data 108 via the communication network 186, and, therefore, cannot be mentally performed.
  • the method 500 allows for parallelism, which allows for encoding (e.g., using first encoder 126-1 of Figure 2B), communicating an encoded byte stream, decoding the encoded byte stream (e.g., using second decoder 126-2 of Figure 2B), or a combination thereof simultaneously using a plurality of graphical data 108 and/or using a plurality of CPUs 172.
  • the present disclosure is not limited thereto.
  • Block 502. referring to block 502, in some embodiments, the method 500 includes obtaining first graphical data 108-1 (e.g., a first image, a set of images, a 3D model, etc.).
  • the first graphical data 108-1 is obtained in electronic form via the communication network 186 and/or retrieved for the memory 192 of the computer system 200. In some embodiments, the first graphical data 108-1 is captured prior to being obtained by the computer system 200. However, the present disclosure is not limited thereto. [00318] In some embodiments, the first one or more graphical data 108-1 is defined by a first plurality of characteristics (e.g., first plurality of graphical data 110-1 of Figure 2A, first plurality of characteristics 110 of method 500 of Figures 5A-5E, etc.).
  • a first plurality of characteristics e.g., first plurality of graphical data 110-1 of Figure 2A, first plurality of characteristics 110 of method 500 of Figures 5A-5E, etc.
  • the first plurality of characteristics 110 includes a sampling resolution of the first graphical data 108, such as a minimum resolution associated with an image in the first graphical data 108, a mode resolution associated with the first graphical data 108, or the like. Moreover, in some embodiments, the first plurality of characteristics 110 further includes a first modality associated with a capture of the first graphical data 108-1, such as a first X-Ray modality or a second thermal imaging modality. [00319] Block 504. Referring to block 504, in some embodiments, the first graphical data 108-1 includes one or more images taken of a first biological sample. In some embodiments, the first graphical data is as exemplified by at least block 406. [00320] Block 506.
  • the first one or more graphical data 108-1 includes one or more images taken of a second biological sample different than the first biological sample.
  • the first graphical data is as exemplified by at least block 408.
  • Block 508 in some embodiments, the sampling resolution of the first one or more graphical data 108-1 includes the minimum resolution associated with the first one or more graphical data 108-1.
  • the first graphical data is as exemplified by at least block 414. DB1/ 147040794.2 81 Attorney Ref. No.: 115834-5037-WO [00322] Block 510.
  • the first modality (e.g., third characteristic 112-3 of Figure 2A) includes a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof.
  • CT computer tomography
  • MRI magnetic resonance imaging
  • PET positron emission tomography
  • radiograph modality a single-photon emission
  • SPE single-photon emission
  • sonography modality e.g., sonography modality
  • the first plurality of characteristics 110 further includes one or more ground truth labels (e.g., fourth characteristic 112-4 of Figure 2A, etc.) with respect to the first modality for the first one or more of graphical data 108-1.
  • the first plurality of characteristics is as exemplified by at least block 424.
  • Block 514 Referring to block 514, in some embodiments, the first plurality of characteristics 110 further includes a respective capacity of the communication network 186, such as a first bandwidth for receiving data through the communication network 186 and/or a second bandwidth for communicating data through the communication network 186.
  • the first plurality of characteristics is as exemplified by at least block 436.
  • Block 516 Referring to block 516, in some embodiments, the first plurality of characteristics 110 further includes a respective capacity of the second computer system 300. In some embodiments, the first plurality of characteristics is as exemplified by at least block 438. [00326] Block 518. Referring to block 518, in some embodiments, the method 500 further includes forming a byte stream map (e.g., byte stream map 600 of Figure 6) for the first one or more graphical data 108-1.
  • a byte stream map e.g., byte stream map 600 of Figure 6
  • the byte stream map 600 includes a plurality of sequence scans (e.g., sequence scans 1010 of method 400 of Figures 4A-4E, scans 1010 of Figure 7, scans 1010 of Figure 8, scans 1010 of Figure 9, scans 1010 of Figure 10, scans 1010 of Figure 11, scans 1010 of Tables 1-14, etc.) through the sampling resolution, which allows an end-user to view, at a highest resolution and/or optimal resolution, the sampling resolution of the first graphical data 108-1 when communicated to a remote device, such as the second computer system 300.
  • each respective sequence scan 1010 in the plurality of sequence scans is associated with a unique rank in a rank order.
  • each rank order defines a sequence of respective sequence scans 1010 in the plurality of sequence scans.
  • the rank order defines DB1/ 147040794.2 82 Attorney Ref. No.: 115834-5037-WO a structured organized set that assigns each unique rank in the rank order to a position in the rank order.
  • the rank order is based on a value of a corresponding resolution.
  • each sequence scan 1010 in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans.
  • each subsequent sequence scan 1010 in the plurality of sequence scans 1010 has an increasing level of quality (e.g., first characteristic 112-1) and resolution (e.g., second characteristic 112-2), such that the complete reconstruction through the final terminal scan 1010 is lossless.
  • the plurality of sequence scans include N scans, in which N is a positive integer, which allows for reconstruction of ⁇ + 1 scans 1010, where the resolution of each scan 1010 is a factor of 2 ⁇ +1 of the original image 108 resolution.
  • each graphical data 108 in the first graphical data is associated with a unique entry or data elements of the byte stream map 600.
  • each sequence scan 1010 is mapped via the byte stream map 6000 to the partial byte stream 114 necessary to decode a corresponding graphical data 108 at the corresponding resolution, or decomposition level, of the sequence scan 1010.
  • an entry, or node, in the byte stream map 600 is created by generating a plurality of tile-part markers in the encoded byte stream 114.
  • each entry, or scan 1010 is associated with at least one marker.
  • the tile- park marker allow for graphical data 108 to be divided into a plurality of tiles, or tessellations, in which each tile is encoded in the byte stream map 600, as opposed to encoded the entirety of the graphical data 108.
  • each tile-part marker is formed in a rectangular shape.
  • each tile park marker is tessellated.
  • the term “tessellated” means a tiling of a plane using one or more geometric shapes, hereinafter “tiles,” with no overlaps or gaps therebetween.
  • the tessellation of a region refers to the tiling of a plane, or surface, that defines the region (e.g., the tessellation of a plane in a 2D polar coordinate system, the tessellation of a surface in a 3D spherical coordinate system).
  • an initial terminal scan 1010 is associated with a first marker and a final terminal scan 1010 is associated with a second marker.
  • the plurality of tile-part markers include a first tile-part marker identified by DB1/ 147040794.2 83 Attorney Ref. No.: 115834-5037-WO byte 0xFF90 and a second tile-part marker identified by byte 0xFFD9 of the encoded byte stream 114, which allows for indicating data in the encoded byte stream for a particular tile- part marker.
  • a number of sequence scans in the plurality of sequence scans is determined as a factor, or function, of 2 ⁇ +1 of the graphical data’s 108 full, or original resolution.
  • the inclusion of tile-part markers in the encoded byte stream map 600 enables graphical data 108 to be communicated at a specific scan 1010 level in the plurality of sequence scans 1010 using a portion of the encoded byte stream 114 without requiring communicating the complete byte stream 114.
  • the byte stream map 600 allows for navigating through the plurality of sequence scans 1010 to access an optimal scan (e.g., block 456 of method 400, block 458 of method 500, etc.) when communicating and/or decoding the encoded byte stream 114 associated with the byte stream map 600.
  • the byte stream map 600 reduces an amount of computational processing required by the first computer system 200 and/or the second computer system 300 by not requiring communication and/or decoding of other scans 1010 in the plurality of sequence scans 1010, such as those that exceed the desired threshold performance.
  • the present disclosure is not limited thereto.
  • the byte stream map 600 allowed for matching tile-part divisions by resolution (e.g., forming the byte stream map 600) in accordance with the tile- part marker for an encoded byte stream 114 of graphical data 108. For instance, in some embodiments, provided that the encoded byte stream 114 to reconstruct the image at a sub- resolution is available, the graphical data is decoded at the scan level and all preceding scan levels even in the absence of the complete encoded byte stream 114 based on the byte stream map 600. This enables quick access to smaller resolution versions of the graphical data (e.g., for thumbnails) based on the byte stream map 600 and the scans 1010 therein, thus minimizing data duplication.
  • resolution e.g., forming the byte stream map 600
  • Block 520 the first one or more graphical data 108 includes 3D graphical data and each scan in the plurality of sequence scans of the 3D graphical data includes a respective 2D layer of the 3D graphical data.
  • the first graphical data is as exemplified by at least block 438.
  • Block 522 Referring to block 522, in some embodiments, the first one or more graphical data 108 includes one or more digital images, one or more digital videos, one or DB1/ 147040794.2 84 Attorney Ref.
  • Block 524 Referring to block 524, in some embodiments, the plurality of sequence scans 1010 includes between 2 scans and 100 scans. In some embodiments, the plurality of sequence scans is as exemplified by at least block 448. [00332] Block 526. Referring to block 526, in some embodiments, the plurality of sequence scans 1010 includes N scans, in which ⁇ is ⁇ ⁇ ⁇ ⁇ 2 ⁇ ⁇ ⁇ ⁇ ⁇ 2 .
  • M is a number of digital assets associated with the second one or more graphical data.
  • ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ 1,2, .. , ⁇ is a native resolution associated with the second one or more graphical data.
  • ⁇ ⁇ ⁇ ⁇ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans.
  • ⁇ is a native resolution associated with the second one or more graphical data
  • ⁇ ⁇ ⁇ ⁇ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans.
  • the plurality of sequence scans is as exemplified by at least block 450.
  • the first plurality of characteristics 110 includes a sampling resolution of the first one or more graphical data 108.
  • the plurality of sequence scans 1010 includes N scans based, at least in part, on the sampling resolution of the first one or more graphical data 108.
  • the first plurality of characteristics is as exemplified by at least block 454. [00339] Block 532.
  • the method 500 includes receiving, from a second computer system 300, a request to evaluate the first one or more graphical data 108 at a first resolution.
  • the first resolution is selected in accordance with a desired threshold performance from performance data.
  • the performance data includes a plurality of training resolutions. In some embodiments, each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, in some such embodiments, each respective training resolution is associated with a corresponding threshold performance for a trained feature extraction model 118 in a plurality of threshold performances. In some embodiments, each threshold performance in the plurality of threshold performances associated with the first modality associated with a capture of a second plurality of graphical data 108. [00340] Block 534. Referring to block 534, in some embodiments, the second plurality of graphical data 108 includes two-dimensional (2D) graphical data, three-dimensional (3D) graphical data, four-dimensional (4D) data, or a combination thereof.
  • the second graphical data is as exemplified by at least block 408.
  • Block 536 in some embodiments, the second plurality of graphical data 108 includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof.
  • the second graphical data is as exemplified by at least block 410.
  • Block 538 Block 538.
  • the second plurality of graphical data 108 includes 50 or more digital images, 100 or more digital images, 1,000 or DB1/ 147040794.2 86 Attorney Ref. No.: 115834-5037-WO more digital images, 10,000 or more digital images, or 100,000 or more digital images.
  • the second graphical data is as exemplified by at least block 412.
  • Block 540 Referring to block 540, in some embodiments, prior to the receiving C), the method 500 further includes training the feature extraction model 118 against the one or more ground truth labels. In some embodiments, the ground truth label and/or training is as exemplified by at least blocks 424-426. [00344] Block 542.
  • the desired threshold performance includes a threshold inter-image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation. In some embodiments, the desired threshold performance is as exemplified by at least block 440. [00345] Block 544.
  • the feature extraction model 118 includes a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof. In some embodiments, the feature extraction model is as exemplified by at least block 420. [00346] Block 546.
  • the feature extraction model 118 includes a neutral network model, a support machine mode, a Na ⁇ ve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph-based model, or a combination thereof.
  • the feature extraction model is as exemplified by at least block 422. [00347] Block 548.
  • the method 500 further includes matching the first resolution to a first sequence scan 1010-1 in the plurality of sequence scans 1010 in accordance with the byte stream map 600.
  • the matching is as exemplified by at least block 456.
  • Block 550 the first codec 124-1 includes a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof.
  • the first codec 124-1 is as exemplified by at least block 444.
  • Block 552 the first codec 124-1 is selected by the second computer system 300.
  • the first codec 124-1 is as exemplified by at least block 446.
  • Block 554 in some embodiments, the first sequence scan 1010-1 includes less than 5% of the encoded byte stream 114. In some embodiments, the first sequence scan is as exemplified by at least block 460. [00351] Block 556. Referring to block 556, in some embodiments, the first sequence scan 1010-1 includes less than 5% of a final terminal scan. In some embodiments, the first sequence scan is as exemplified by at least block 462. [00352] Block 558.
  • the method 500 includes encoding, in accordance with a corresponding encoder 126 of a first codec 124-1, the first one or more graphical data 108, thereby forming an encoded byte stream 114.
  • the encoded byte stream 114 includes the plurality of sequence scans through at least the first resolution.
  • the encoded byte stream 114 is as exemplified by at least block 442. [00353] Block 560.
  • the encoded byte stream 114 is associated with a first file size of the final terminal scan 1010 in the plurality of sequence scans, and a second file size of the first sequence scan 1010 in the plurality of sequence scans is less than the first file size of the final terminal scan 1010.
  • the encoded byte stream 114 is as exemplified by at least block 464.
  • Block 562 in some embodiments, the method 500 further includes communicating, via a communication network 1016, in accordance with the first codec 124-1, the encoded byte stream 114 and one or more instructions to terminate decoding of the encoded byte stream 114 at the second computer system 300 to the second computer system 300.
  • the one or more instructions to terminal decoding execute when the decoding decodes the first sequence scan 1010 in the plurality of sequence scans at the second computer system 300, thereby optimizing decoding of graphical data.
  • the communicating is as exemplified by at least block 458.
  • the first experiment employed a 2D classification model (e.g., first model 118-1 of Figure 2B) that detected abnormalities in a second one or more graphical data (e.g., graphical data 108-2 of Figure 2A) associated with a first modality (e.g., first DB1/ 147040794.2 88 Attorney Ref. No.: 115834-5037-WO characteristic 112-1) of the thorax using chest x-ray (CXR) graphical data.
  • a 2D classification model e.g., first model 118-1 of Figure 2B
  • graphical data e.g., graphical data 108-2 of Figure 2A
  • a first modality e.g., first DB1/ 147040794.2 88 Attorney Ref. No.: 115834-5037-WO characteristic 112-1
  • the second experiment employed a 3D segmentation model (e.g., second model 118-2 of Figure 2B) that segmented a third plurality of graphical data (e.g., graphical data 108-3 of Figure 2A) associated with a second modality (e.g., second characteristic 112-2) of the liver and spleen using computed tomography (CT) data.
  • a 3D segmentation model e.g., second model 118-2 of Figure 2B
  • a third plurality of graphical data e.g., graphical data 108-3 of Figure 2A
  • a second modality e.g., second characteristic 112-2
  • the models 118 were trained using a first plurality of graphical data (e.g., graphical data 108-1 of Figure 2A) that was local to a first computer system (e.g., graphical data system 200 of Figures 2A and 2B), validated using held-out internal test sets, and evaluated on unseen (e.g., discrete from the first plurality of graphical data) graphical data received from (e.g., via communication network 186) a hospital system (e.g., client device 300 of Figure 3).
  • a first computer system e.g., graphical data system 200 of Figures 2A and 2B
  • unseen e.g., discrete from the first plurality of graphical data
  • the graphical data 106 communicated, decode time using a decoder 128, and throughput (e.g., a number of images processed by a respective model 118 per unit time (e.g., second)), which were measured as data metrics and compared to an original non-high throughput JPEG 2000 (HTJ2K) version of the respective plurality of graphical data 108.
  • throughput e.g., a number of images processed by a respective model 118 per unit time (e.g., second)
  • a trained multi-label DenseNet121 model 118 architecture was utilized, initialized with ImageNet weights, on the NIH Chest X-Ray (“NIH”) dataset 108 (e.g., first graphical data of block 404 of method 400, second graphical data of block 532 of method 500, etc.) using transfer learning. See Huang et al., 2017, “Densely Connected Convolutional Networks,” Proceedings of the IEEE conference on computer vision and pattern recognition.
  • the NIH graphical data 108 included 14 disease labels (e.g., ground truth label characteristics 112) with 112,120 frontal-view CXRs (e.g.g., chest x-ray modality characteristic 112) from 30,805 patients (e.g., first graphical data 108-1 of Figure 2A, first graphical data 108 of block 404 of method 400 of Figures 4A-4E, etc).
  • Graphical data 108 within the held-out NIH test set were originally encoded as a PNG filetype (e.g., first PNG code 124-1 using first encoder 126-1, first codec 124-1 of block 442 of method 400 DB1/ 147040794.2 89 Attorney Ref. No.: 115834-5037-WO of Figure 4A-4E, first code 124-1 of block 558 of method 500 of Figures 5A-5E, etc.) with a normalized resolution of 1024x1024.
  • PNG filetype e.g., first PNG code 124-1 using first encoder 126-1, first codec 124-1 of block 442 of method 400 DB1/ 147040794.2 89 Attorney Ref. No.: 115834-5037-WO of Figure 4A-4E, first code 124-1 of block 558 of method 500 of Figures 5A-5E, etc.
  • the NIH test set 108 was progressively encoded 128 using lossless HTJ2K codec (e.g., second codec 124-1 of Figure 2B), which included 5 scans 1010 in a plurality of sequence scans that yielded a maximum compression ratio of 256:1 for the corresponding byte stream map 600 (e.g., byte stream map 60 of method 500 of Figures 5A-5E).
  • a first resolution associated with an optimal scan 1010 in the plurality of sequence scans was identified (e.g., matching of block 456 of method 500, matching of block 548 of method 500, etc.) and evaluated the model 118 across each scan 1010 of HTJ2K encoded byte stream 114 and original PNG graphical data 108.
  • the MIMIC test set was progressively encoded 126 using lossless HTJ2K code 124, which included 7 scans 1010 in a plurality of sequence scans, to yield a maximum compression ratio of 4096:1.
  • Performance of a respective model 118 was evaluated across each scan 1010 of the HTJ2K encoded byte stream 114 and original JPEG data 108 by a desired threshold performance using a binary cross entropy loss (BCE) (e.g., a first threshold performance criterion), an area under the receiver operating characteristic 112 (AUROC) (e.g., a second desired threshold performance), and an area under the precision recall curve (AUPRC) (e.g., a third threshold performance criterion).
  • BCE binary cross entropy loss
  • AUROC area under the receiver operating characteristic 112
  • AUPRC area under the precision recall curve
  • the AUROC and AUPRC scores 112 were defined as the mean score across all disease labels.
  • a 95% confidence interval (CI) for the AUROC scores was determined and compared using bootstrapping and a paired two-tailed t-test (e.g., third model 118-3 of Figure 2B). Statistical significance was defined as p ⁇ 0.05. DB1/ 147040794.2 90 Attorney Ref.
  • a trained 3D-UNet model 118 used the MONAI framework for segmentation of liver and spleen graphical data 108 and used the corresponding graphical datasets 108 from the Medical Segmentation Decathlon (MSD) training sets (e.g., first graphical data of block 404 of method 400, second graphical data of block 532 of method 500, training data of Example 3, etc.).
  • MSD Medical Segmentation Decathlon
  • the model 118 architecture included 5 two-convolutional residual units with down/up-sampling by a factor of 2 at each unit. All graphical data 108 volumes were resized to a resolution of 256x256x128 and voxel values normalized between 0 and 1 Random. Foreground patches of resolution 128x128x32 were extracted from each volume such that a center voxel of each patch belonged to either a foreground class or a background class.
  • Graphical data 108 within the held-out MSD liver and spleen test sets were originally encoded as NifTI 124 volumes with a normalized resolution of 512x512xN.
  • One limitation of the HTJ2K standard 124 is its inability to encode 3D graphical data 108.
  • 3D graphical data was progressively encoded by each 2D slice using lossless HTJ2K 124, which yielded four scans 1010 in a plurality of sequence scans, which yielded a maximum compression ratio of 64:1.
  • All slices were decoded (e.g., using a first decoder 128-1 of Figure 2B) individually, reconstructed into a volume, and evaluated with the model 118, along with the original NifTI volumes.
  • the scan 3 in the plurality of sequence scans yielded excellent DICE scores of 0.88 ⁇ 0.09 and 0.81 ⁇ 0.17 on liver and spleen, respectively, which required 90.32% less data to communicate, 90.26% less time to decode, and yielded a factor of 1.61 increase in throughput.
  • lossless HTJ2K volumes of graphical data 108 required 73.12% less data to communicate and 81.20% less time to decode when compared to the original NifTI 124 compressed volumes.
  • Example 2 [00372] In this Example, encoding parameters (e.g., a number of scans in a plurality of sequence scans) were determined by analyzing a plurality of graphical data (e.g., first graphical data 108-1 of Figure 2A, second graphical data 108-2 of Figure 2A, graphical data of Example 1, graphical data of Example 3, etc.). A number of scans 1010, N, in the plurality of sequence scans was determined by evaluating the entirety of a respective plurality of graphical data (e.g., the entire second one or more graphical data 108-2), such that ⁇ was log 2 ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ 2 .
  • a number of scans in a plurality of sequence scans were determined by analyzing the entirety of a respective plurality of graphical data (e.g., the entire second one or more graphical data 108-2), such that ⁇ was log 2 ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ 2 .
  • ⁇ was a native resolution of the graphical data
  • ⁇ was the number of graphical data within the set
  • ⁇ ⁇ ⁇ ⁇ was the minimum resolution of the first scan.
  • the optimal set of scans 1010 to encode for each graphical data within the second one or more graphical data was determined, which ensured the resolution of the first scan (e.g., an initial terminal scan in the plurality of sequence scans) satisfied a threshold ⁇ ⁇ ⁇ ⁇ , such as the initial terminal scan exceeded ⁇ ⁇ ⁇ ⁇ .
  • the encoded byte stream encoded by the code 124 ensured normalization of each resolution associated with a corresponding scan in the plurality of sequence scans, which included the initial terminal sequence scan and any subsequent scan to prevent resolution disparities.
  • the graphical data was encoded 126 using fewer scans 1010, while graphical data 108 with higher resolutions using more scans.
  • a first image with resolution 1470x1461 in the second one or more graphic data was encoded with 6 scans in a first plurality of sequence scans with an initial terminal scan with resolution of 46x46.
  • a second image with resolution 4280x3520 in the second one or more graphic data was encoded with 7 scans in a second plurality of DB1/ 147040794.2 94 Attorney Ref. No.: 115834-5037-WO sequence scans with an initial terminal scan with resolution of 67x55.
  • This approach ensured that all images within the second one or more graphical data yielded an initial terminal scan with a resolution within a range of a corresponding resolution of the initial terminal scan of the first and second images, such as range of [ ⁇ 45, ⁇ 90].
  • the corresponding resolution of all subsequent scans in the plurality of sequence scans was normalized with a fixed range of resolutions.
  • a progressive encoding, decoding, and communication framework was based on lossless progressive encoding, a type of compression offered by the JPEG 2000 standard (e.g., first codec 124-1 of Figure 2B, first code 124-1 of Figure 3, first codec 124-1 of block 422 of method 400 of Figures 4A through 4E, first codec 124-1 of block 558 of method 500 of Figures 5A through 5E, etc.) to optimize the delivery of images (e.g., first graphical data 108-1 of Figure 2A, graphical data 108-1 of block 404 of method 400 of Figures 4A through 4E, graphical data 108-1 of block 502 of method 500 of Figures 5A through 5E, etc.) over the internet (e.g., communication network 186 of Figure 2A, communication network 186 of Figure 3, communication network 186
  • a low-resolution image (e.g., second scan 1010-2 of Figure 11) was displayed immediately and gradually increased in resolution over time.
  • Progressive encoding 126 was performed by encoding a medical image (e.g., graphical data 108) as a series of lossy compressed decompositions (e.g., a plurality of sequence scans 1010), where each subsequent sequence scan 1010 had an increasing level of quality and resolution, such that the complete reconstruction was lossless at the final terminal scan 0110.
  • a chest x-ray (“CXR”) with a resolution of 512x512 was progressively encoded 126 with three scan 1010 levels (e.g., three sequence scans in the plurality of sequence scans 1010) and then decoded at sub-resolutions 64x64 for an initial terminal scan, 128x128 for a DB1/ 147040794.2 95 Attorney Ref.
  • the chest x-ray (“CXR”) with a resolution of 512x512 was progressively encoded 126 with three scan 1010 levels then decoded at sub-resolutions 64x64 for an initial terminal scan, 128x128 for a second scan, and 256x256 for the final terminal scan as well as at the original resolution 512x512.
  • High-Throughput JPEG 2000 (HTJ2K) [00380] Recently, increased popularity of High-Throughput JPEG 2000 (HTJ2K, second code 124-2 of Figure 2B, second code 124-2 of Figure 3, etc.) renewed interest in progressive encoding within the medical imaging community with its integration into the DICOM transfer syntax and cloud-based imaging infrastructures like Amazon Web Services’ HealthImaging.
  • HTJ2K is an extension of the JPEG 2000 standard with significantly faster encoding and decoding at the expense of a slight reduction in encoding efficiency. For lossless encoding and decoding, HTJ2K was more than 30 times faster than JPEG 2000 – putting it on par with the heavily optimized JPEG standard in both single and multi-threaded applications.
  • Example 4 Development of an Intelligent Streaming Framework for Reconstruction of Medical Images at Target Resolution
  • Intelligent Streaming [00383]
  • This Example discloses the intelligent streaming and subsequent decoding of partial byte streams 114 to reconstruct progressively encoded medical images 108 at an optimal resolution associated with a scan 1010 for AI model 118 inference (e.g., graphical data system 200 of Figure 1, graphical 200 of Figures 2A and 2B, client device 300 of Figure 1, client device 300 of Figure 3, second computer system block 428 of Figure 4C, second computer system of block 532 of method 500, etc.).
  • AI model 118 inference e.g., graphical data system 200 of Figure 1, graphical 200 of Figures 2A and 2B, client device 300 of Figure 1, client device 300 of Figure 3, second computer system block 428 of Figure 4C, second computer system of block 532 of method 500, etc.
  • the optimal resolution (e.g., first characteristic 112-1 of Figure 2A, first training resolution of block 456 of method 400 of Figures 4A-4E, first resolution of block 548 of method 500 of Figures 5A-5E, etc.) was matched based on the smallest decomposition level (e.g., sequence scan 1010) at which the reconstructed medical image 108 contained sufficient information for an AI system (e.g., first model 118-1 of Figure 2A, first model 118-1 of Figure 3, trained feature extraction model of block 418 of method 400 of Figures 4A-4E, trained feature extraction model of block 532 of method 500 of Figures 5A-5E, etc.) to accurately characterize the image (e.g., block 532 of Figure 5C, block 456 of Figure 4E, etc.).
  • an AI system e.g., first model 118-1 of Figure 2A, first model 118-1 of Figure 3, trained feature extraction model of block 418 of method 400 of Figures 4A-4E, trained feature extraction model of block 532 of method 500 of Figures
  • FIG. 16 An automated system that intelligently determined and validated the optimal resolution associated with a scan 1010 for streaming medical images 108 to an AI system 118 for inference without impacting its diagnostic performance (Figure 16), which comprised of three primary components: a progressive encoder, a stream optimizer, and a progressive decoder.
  • Figure 16 An automated system that intelligently determined and validated the optimal resolution associated with a scan 1010 for streaming medical images 108 to an AI system 118 for inference without impacting its diagnostic performance (Figure 16), which comprised of three primary components: a progressive encoder, a stream optimizer, and a progressive decoder.
  • Progressive Encoder [00385] A progressive encoder as described in Example 3 was obtained.
  • the role of the progressive encoder was to encode medical images into byte streams using HJT2K ( Figure 16).
  • OpenJPHpy version 0.1
  • a Python library for OpenJPH an open-source implementation of HTJ2K
  • ⁇ decomposition levels e.g., N scans in the plurality of scans
  • tile-part marker divisions by scan e.g., N scans in the plurality of scans
  • tile-part markers e.g., N scans in the plurality of scans
  • the encoding of 3D imaging data 108 required each slice of the volume to be encoded separately as a 2D image. This was particularly applicable to encoding medical images 108 stored in the NIfTI format 124 – which encoded the entire volume in a single file. Furthermore, the rescale intercept and slope were calculated to rescale pixel values to 16-bit unsigned integers when required prior to encoding a medical image (e.g., when encoding Hounsfield units for CT scans).
  • a stream optimizer was obtained that enabled the streaming and subsequent decoding of partial byte streams 114 to reconstruct medical images at the scan 1010 corresponding to the optimal resolution for the AI system 118.
  • the goal of the stream optimizer was to provide an interface for an AI system 300 to access the database of medical images 106 for inference from the healthcare provider 200.
  • the stream optimizer included two components: a host-side stream optimizer, and a client-side stream optimizer (Figure 16).
  • the role of the host-side stream optimizer was to interface with the database of progressively encoded medical images (e.g., graphical data store 106 of Figure 2A, etc.) obtained using the progressive encoder described herein and in Example 3, to build a stream optimization map (e.g., byte stream map of 518 of method 500 of Figures 5A-5E, etc.).
  • a stream optimization map e.g., byte stream map of 518 of method 500 of Figures 5A-5E, etc.
  • Each medical image 108 included a unique entry in the stream byte map 600.
  • the stream by ten map 600 allowed for mapping each sub-resolution to the partial byte stream necessary to decode the image at the sub-resolution’s corresponding decomposition level.
  • An entry was DB1/ 147040794.2 98 Attorney Ref.
  • No.: 115834-5037-WO created by locating tile-part markers in the complete byte stream (identified by bytes 0xFF90 and 0xFFD9) and calculating the sub-resolution as a factor of 2 ⁇ +1 of the image’s full resolution.
  • the inclusion of tile-part markers in the encoding step allowed for building of the stream optimization map. This enabled a medical image to be streamed at a specific decomposition level using partial byte streams without streaming the complete byte stream. For example, a CXR with resolution 512x512 was streamed at sub-resolution 256x256 with the first 12 KB of the complete 180 KB byte stream.
  • the role of the client-side stream optimizer was to interface with the AI system to intelligently determine and validate the optimal resolution for streaming medical images from the host without impacting the AI system’s diagnostic performance.
  • the client- side stream optimizer used the AI vendor’s internal test set to determine the optimal resolution for the AI system.
  • the internal test set was first progressively encoded, then decoded at each decomposition level, and finally evaluated with the AI system to measure performance metrics (e.g., AUROC for classification).
  • the optimal resolution was determined as the smallest decomposition level which corresponded to a sub-resolution closest to the input resolution of the AI system and resulted in no statistically significant difference in the AI system’s performance compared to the full resolution test set.
  • Example 5 Generating Datasets for use in Intelligent Streaming Framework DB1/ 147040794.2 99 Attorney Ref. No.: 115834-5037-WO
  • six medical imaging datasets were obtained, each dataset including graphical data in the form of medical images. Three datasets were obtained for each of (i) 2D classification with CXRs and (ii) 3D segmentation with CT scans.
  • NIH National Institutes of Health ChestX-ray14 dataset
  • All images in the dataset had a fixed resolution of 1024x1024, had pixel values normalized to range 0–255, and were encoded in the PNG format (e.g., lossless).
  • Seven disease labels from the NIH dataset were considered for analysis: Atelectasis, Cardiomegaly, Consolidation, Edema, Pleural Effusion, Pneumonia, and Pneumothorax.
  • MIMIC MIMIC
  • Images in the dataset had a mean resolution of 2500x3056, had pixel values normalized with histogram equalization to range 0–255, and were encoded in the JPEG format (e.g., lossy).
  • the MIMIC-CXR dataset provides the images in the 12-bit uncompressed DICOM Little Endian format.
  • the same preprocessing steps, excluding JPEG conversion, were applied to the DICOM images during inference.
  • the same seven disease labels as the NIH dataset were considered for analysis.
  • Example 6 Assay Development for Classification and Segmentation of Graphical Data using an Intelligent Streaming Framework [00405] Experimental Design [00406] AI inference 118 was evaluated across two experiments: 2D classification of abnormalities with CXRs, and 3D segmentation of the liver and spleen organs with abdomen CT scans using the datasets for each task described in Example 5.
  • a deep learning (DL) model 118 (in accordance with the framework described in Examples 3 and 4) was trained using an internal dataset (e.g., a training data set comprising a first plurality of graphical data, in accordance with the present disclosure), the optimal resolution for AI inference was determined and validated using an internal test set from the same site, and the optimal resolution was then evaluated on external data streamed from a different site (e.g., a second one or more graphical data, in accordance with the present disclosure). [00407] For each test set, the mean structural similarity index measure (SSIM) and mean peak signal-to-noise ratio (PSNR) were measured as image fidelity metrics.
  • SSIM structural similarity index measure
  • PSNR mean peak signal-to-noise ratio
  • the amount of transferred data, decoding time, inference time, and throughput were measured as data efficiency metrics. Both sets of metrics were measured at each decomposition level (including the optimal resolution) for the Intelligent Streaming framework (“ISLE”) streamed version and compared with the original non-ISLE streamed versions of the test set.
  • the SSIM and PSNR after normalizing pixel values to range 0–1 were measured using scikit- image (version 0.22).
  • the decode time was measured using OpenCV (version 4.7) for JPEG and PNG formats, Pydicom (version 2.3.1) for uncompressed DICOM imaging data, and OpenJPHpy (version 0.1) for HTJ2K format.
  • the inference time was measured by the time taken by the AI system to characterize a medical image, including any preprocessing steps such as resizing to input resolution. Both decoding and inference times were measured on a DB1/ 147040794.2 101 Attorney Ref. No.: 115834-5037-WO single thread, without parallel processing.
  • a multi-label DenseNet121 trained model 118 pre-trained with ImageNet, using transfer learning for 100 epochs with binary cross-entropy loss, batch size of 64, learning rate of 5e-5, and early stopping. All CXRs were resized to resolution 224x224 while maintaining aspect ratio with zero-padding with pixel values normalized to range 0–1 and scaled to ImageNet statistics. Random augmentations (e.g., rotation, flip, zoom, contrast, or a combination thereof) were applied to input images during training. Then, the internal NIH test set was used to determine and validate the optimal resolution. Finally, the optimal resolution on the external CheXpert and MIMIC-CXR-JPG datasets was evaluated.
  • a modified 3D- UNet model 118 was trained with five levels of two residual convolutional units on the internal MSD Liver dataset for 500 epochs with Dice loss, batch size of 2, decaying learning rate with cosine annealing scheduler, and initial learning rate of 1e-4. All CT scans were resized to resolution 256x256x128, voxel spacing rescaled to 1.5x1.5x2.0 mm, and pixel values normalized to range 0–1. Random foreground patches of resolution 128x128x32 were extracted from each volume and centered on voxels belonging either to the foreground or background class. Random augmentations (e.g., rotation, scaling, elastic deformation, or a combination thereof) were applied to input volumes during training.
  • Random augmentations e.g., rotation, scaling, elastic deformation, or a combination thereof
  • the internal MSD Liver test set was used to determine and validate the optimal resolution using our framework. Finally, the optimal resolution was determined on the external BTCV Liver dataset.
  • DB1/ 147040794.2 102 Attorney Ref. No.: 115834-5037-WO [00412]
  • Example 7 Intelligent Streaming Framework Optimizes Decoding of Graphical Data for Classification Tasks by Determining Optimal Resolution of Images
  • Results – Classification [00417] Using the internal NIH test set described in Example 5 and according to the assay parameters described in Example 6, the optimal resolution for the DL model was determined as the third decomposition level with resolution 256x256, which was the closest resolution to the model’s input size of 224x224.
  • the encoded byte stream 114 required 91.58% less data to be transferred, 87.65% less time to decode, and 11.61% less time for inference while increasing throughput by 3.72x.
  • ISLE required 99.83% less data to be transferred, 99.53% less time to decode, and 97.23% less time for inference while increasing the throughput by 113.87x.
  • the ISLE framework only required 98.62% less data to be transmitted, 98.49% less time to decode, and 18.73% less time for inference while increasing throughput by 30.12x.
  • Example 8 Intelligent Streaming Framework Optimizes Decoding of Graphical Data for Segmentation Tasks by Determining Optimal Resolution of Images [00421] Results – Segmentation [00422] Liver Segmentation [00423] Using the internal MSD Liver test set described in Example 5 and according to the assay parameters described in Example 6, the ISLE framework determined the optimal resolution of the DL model as the third decomposition with resolution 256x256 and variable volumes sizes, which was the closest resolution to the model’s input size of 256x256x128.
  • the ISLE framework required 81.58% less data to be transferred, 87.86% less time to decode, and 2.03% less time for inference while increasing the throughput by 3.42x.
  • the encoded byte stream 114 takes further enabled healthcare providers and AI vendors to benefit from the inherent gains in data efficiency from streaming a lower resolution image.
  • the uncompressed external MIMIC test set shrunk almost 600 times in size – from 199 GB to 0.34 GB – when the medical images were progressively encoded using HTJ2K and then streamed at the optimal resolution.
  • the streaming of medical images at the optimal resolution also improved the computational efficiency of AI inference – a key factor for reducing the carbon footprint of AI deployments.
  • the encoded byte stream 114 advantageously allows for the maintenance of equity in care across both small and large healthcare systems by dramatically reducing the infrastructural requirements for healthcare providers.
  • Examples 3-8 indicate that the benefits of the encoded byte stream 114 are not just dependent on the task and modality, but also the architecture of the AI system.
  • a DL segmentation model 118 with CT scans trained on standard acquisition resolutions yielded significantly fewer gains in data and computational efficiency compared to a model DB1/ 147040794.2 112 Attorney Ref. No.: 115834-5037-WO trained on lower resolutions.
  • the encoded byte stream 114 did not perform 3D compression and each axial slice of the volume was encoded and decoded as a separate image. This added an additional step to reconstruct the volume after decoding.
  • the encoded byte stream 114 utilized medical images progressively encoded using either JPEG 2000 or HTJ2K for intelligent streaming. This was allowed existing imaging infrastructures that do not use compression to store medical images.
  • Examples 3-8 explored training DL models such that they were robust on inference with low resolution images for medical image AI applications on the edge through techniques like data augmentation. These Examples explored the potential for reducing the amount of data storage and transmission for cloud-based imaging databases with a focus on greater accessibility and interoperability of large-scale medical images datasets for AI applications. [00439] These Examples presented an important first step towards a framework for intelligently streaming medical images to AI vendors at an optimal resolution for inference.
  • the encoded byte stream 114 addressed inefficiencies in current imaging infrastructures by improving data and computational efficiency of AI deployments in the clinical environment without impacting clinical decision-making using AI systems.
  • the present invention can be implemented as a computer program product that includes a computer program mechanism embedded in a non-transitory computer-readable storage medium.
  • the computer program product could contain instructions for operating the user interfaces disclosed herein.
  • These program modules can be stored on a CD-ROM, DVD, magnetic disk storage product, USB key, or any other non-transitory computer readable data or program storage product.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Medical Informatics (AREA)
  • Public Health (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Epidemiology (AREA)
  • Data Mining & Analysis (AREA)
  • Primary Health Care (AREA)
  • Databases & Information Systems (AREA)
  • Pathology (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Radiology & Medical Imaging (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)

Abstract

Systems and methods for optimizing encoding, communicating, decoding, or a combination thereof of graphical data are provided. A training data set including first graphical data defined by a sampling resolution and first modality are obtained. A trained feature extraction model and performance data are obtained. Second graphical data is encoded forming an encoded byte stream including a plurality of sequence scans. Each scan is associated with a unique rank order defining a sequence and a corresponding progressive resolution that is based on a resolution of a preceding scan or a predetermined resolution. A first training resolution is matched to the desired threshold performance. The encoded byte stream and instructions to terminate decoding of the encoded byte stream at the second computer system when the decoding decodes a sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution are communicated.

Description

Attorney Ref. No.: 115834-5037-WO SYSTEMS AND METHODS FOR HIGH-THROUGHPUT ANALYSIS FOR GRAPHICAL DATA CROSS-REFERENCE TO RELATED APPLICATION [0001] The present Application claims priority to United States Patent Application No.: 63/501,552, entitled “Systems and Methods for High-Throughput Analysis for Graphical Data,” filed May 11, 2023, which is hereby incorporated by reference in its entirety for all purposes. TECHNICAL FIELD [0002] The present disclosure relates to systems and methods for high-throughput analysis for graphical data. More particularly, the present disclosure relates to systems and methods for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof. BACKGROUND [0003] The utility of artificial intelligence (AI) systems in medical imaging has grown rapidly due in part to the curation of large-scale datasets containing two-dimensional (2D) images and three-dimensional (3D) volumes across various image modalities. Using such datasets, AI systems have the ability to learn complex relationships, such as between different tissue types to provide the capability to detect and/or predict the presence of disease. As a result, entities have begun to develop and deploy intelligent imaging solutions in hospital systems across a series of different clinical applications, such as Amazon’s HealthLake and Google’s Medical Imaging Suite. See Amazon Healthlake, available at aws.amazon.com/healthlake/ (accessed February 27, 2023); Google Medical Imaging Suite, available at cloud.google.com/medical-imaging, (accessed February 27, 2023), each of which is hereby incorporated by reference in its entirety for all purposes. [0004] Deploying intelligent imaging solutions typically involves healthcare providers streaming large volumes of medical image data to AI vendors for analysis. However, limitations in bandwidth create communication bottlenecks when streaming medical images DB1/ 147040794.2 1 Attorney Ref. No.: 115834-5037-WO over the internet, leading to delays in patient diagnosis and treatment. Furthermore, as the adoption of AI systems in the clinical setup continues to increase, healthcare providers and AI vendors will require greater computational infrastructure, therefore dramatically increasing costs. As smaller family practices and rural clinics without high-quality computational infrastructure will struggle to keep pace with larger hospital systems, health disparities in rural populations will undoubtedly widen. See Douthit et al., 2015, “Exposing Some Important Barriers to Health Car Access in the Rural USA,” Public Health, 129(6), pg. 611-620, which is hereby incorporated by reference in its entirety for all purposes. [0005] Progressive encoding has revolutionized the way media content is consumed over the internet. Prior to the advent of progressive encoding, entire media files had to be downloaded before being viewable to an end-user. This process of requiring downloading entire media files is often slow and inefficient, especially for end-users with slow internet connections or limited storage and memory on their devices. [0006] Progressive encoding works by encoding the imaging data into a series of progressive scans, each with an increasing level of detail. These techniques allow for the efficient delivery of large datasets in real-time and have proven useful in enhancing user experience in several applications, including video streaming and web applications. However, the utility of progressive encoding in optimizing the AI inference workflow for clinical decision making at scale has not yet been explored. [0007] Conventional approaches for encoding large sets of graphical data follows a one-size- fits-all approach. Even though this one-size-fits-all is sufficiently versatile for most applications, this conventional approach for encoding is deficient when needing to transmit and/or decode large amounts of data without noticeable latency. For example, media streaming entities require transferring hundreds of gigabytes of data to each viewer device in order to deliver a high-resolution (e.g., 4K) video stream. However, using conventional encoding techniques results in greater infrastructure costs, such as due to bandwidth and storage constraints. Moreover, using conventional encoding techniques results in an inferior experience for the viewer, such as due to an increased time required to decode a transmitted encoding and display each video frame. [0008] Given the above background, what is needed in the art are improved systems and methods for providing progressive encoding of graphical data that enables high throughput, DB1/ 147040794.2 2 Attorney Ref. No.: 115834-5037-WO bandwidth optimized, and computationally efficient analysis with various computational models. SUMMARY [0009] The present disclosure addresses the shortcomings disclosed above by providing systems and methods for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof via a communication network. More particularly, the systems and methods of the present disclosure train a feature extraction model using a first plurality of graphical data, encode a second one or more graphical data forming an encoded byte stream, match a desired performance to a first training resolution, communicate an optimal subset of an encoded byte stream, terminate communication of the encoded byte stream, or a combination thereof, which allows for less time to complete a process such as two-dimensional (2D) graphical data, three-dimensional (3D) graphical data, four-dimensional (4D) graphical data, or a combination thereof, such as for application with the feature extraction model to complete tasks of classification and/or segmentation across different subject matter (e.g., different organs, tissues, bones, etc.) and imaging modalities. In some embodiments, the systems and methods of the present disclosure encode the second one or more graphical data based on a plurality of characteristics that, collectively, define the second one or more graphical, which allows for course-grain utilizations with the second one or more graphical data. In some embodiments, the systems and methods of the present disclosure encode the second one or more graphical data based on a subset of characteristics in the plurality of characteristics that is specific to a respective graphical data in the second one or more graphical data, which allows for a fine-grain utilizations with the second one or more graphical data set. In some embodiments, moreover, the systems and methods of the present disclosure result in faster turnaround times when processing a request for an evaluation of various graphical data, and reduced overall cost of graphical data storage and communication, without negatively impacting decision making using computational models and one or more computer systems. [0010] Accordingly, in some embodiments, the system and methods of the present disclosure are utilized match an optimal subset of the encoded byte stream in the form a sequence scan in a plurality of sequence scans of the encoded byte stream that is required for utilization with computer-implemented models, communicate the optimal subset of the encoded byte stream DB1/ 147040794.2 3 Attorney Ref. No.: 115834-5037-WO via a communication network, efficiently reconstruct original graphical data from the optimal subset of the encoded byte stream, or a combination thereof. [0011] In some embodiments, the systems and methods of the present disclosure enable an end-user, such as a clinician at a second computer system, which includes radiologists, pathologists, oncologists, or the like, to receive a similarly optimized subset of the encoded byte stream, such as for utilizing with high-throughput clinical decision making and diagnosis. Accordingly, in some embodiments, the systems and methods of the present disclosure match an optimal resolution for a plurality of graphical data of an encoded byte stream based on, for instance, a clinical use case and/or a form factor (e.g., hardware specifications) of a device including one or more feature extraction models configured to perform an evaluation on a respective modality of graphical data. As such, the systems and methods of the present disclosure yield higher throughput, without negatively impacting clinical decision making as well as performance when using a respective computational model with the encoded byte stream, resulting in faster turnaround times, and reduced overall cost of data storage and transmission in comparison to conventional progressive encoding and/or decoding techniques. [0012] Turning to more specific aspects, one aspect of the present disclosure is directed to providing a method for optimizing decoding of graphical data. The method includes using a first computer system. The first computer system includes a processor and memory. In some embodiments, the method includes obtaining, in electronic form, a training data set. The training data set includes a first plurality of graphical data. The first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data. Moreover, the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data. In addition, in some embodiments, the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data. The performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances. Moreover, the method includes receiving a request to evaluation a second one or more graphical data at a second computer system. Moreover, the second one or more graphical data is defined by a second plurality of DB1/ 147040794.2 4 Attorney Ref. No.: 115834-5037-WO characteristics. The second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data. Furthermore, the second plurality of characteristics includes a desired threshold performance. The method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. Furthermore, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. Furthermore, the method includes matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data. Additionally, the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream at the second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, thereby optimizing decoding of graphical. [0013] In some embodiments, the first plurality of graphical data includes one or more images taken of a first biological sample. [0014] In some embodiments, the second one or more graphical data includes one or more images taken of a second biological sample different than the first biological sample. [0015] In some embodiments, the first plurality of graphical data includes as 2D graphical data, 3D graphical data, 4D graphical data, or a combination thereof. [0016] In some embodiments, the second one or more graphical data includes 3D graphical data. Moreover, each scan in the plurality of sequence scans of the 3D graphical data includes a respective 2D layer of the 3D graphical data. [0017] In some embodiments, the first plurality of graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. DB1/ 147040794.2 5 Attorney Ref. No.: 115834-5037-WO [0018] In some embodiments, the first plurality of graphical data includes 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images. [0019] In some embodiments, the second one or more graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non- fungible token assets, or a combination thereof. [0020] In some embodiments, the sampling resolution of the first plurality of graphical data includes the minimum resolution associated with the first plurality of graphical data. [0021] In some embodiments, the first modality includes a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof. [0022] In some embodiments, the first codec includes a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof. [0023] In some embodiments, the first codec is selected by the first computer system. [0024] In some embodiments, the plurality of sequence scans includes between 2 scans and 100 scans. [0025] In some embodiments, the plurality of sequence scans includes N scans, in which In some embodiments, M is a number of digital assets associated with
Figure imgf000008_0001
the second one or more graphical data. In some embodiments, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data. Moreover, in some embodiments, ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. [0026] In some embodiments, the plurality of sequence scans includes N scans, in which ^^^^ = ^^^^ ^^^^ ^^^^ ∑ ^^^^ ^^^^=1 ^^^^ ^^^^ 2� ^^^^ ^^^^ ^^^^ ^^^^ ^^^^� − 1, M is a number of digital assets associated with the second one or more graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. DB1/ 147040794.2 6 Attorney Ref. No.: 115834-5037-WO [0027] In some embodiments, the first plurality of characteristics further includes one or more ground truth labels with respect to the first modality for the first plurality of graphical data. [0028] In some embodiments, the using the first plurality of graphical data further includes training the feature extraction model against the one or more ground truth labels. [0029] In some embodiments, the second plurality of characteristics includes a sampling resolution of the second one or more graphical data. Moreover, in some embodiments, the plurality of sequence scans includes N scans that is based, at least in part, on the sampling resolution of the second one or more graphical data. [0030] In some embodiments, the second plurality of characteristics further includes a respective capacity of the communication network. [0031] In some embodiments, the second plurality of characteristics further includes a respective capacity of the first computer system. [0032] In some embodiments, the feature extraction model includes a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof. [0033] In some embodiments, the feature extraction model incudes a neutral network model, a support machine mode, a Naïve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph-based model, or a combination thereof. [0034] In some embodiments, the first sequence scan includes less than 5% of the encoded byte stream. [0035] In some embodiments, the first sequence scan includes less than 5% of the final terminal scan. [0036] In some embodiments, the desired threshold performance includes a threshold inter- image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation. [0037] In some embodiments, the encoded byte stream is associated with a first file size of the final terminal scan in the plurality of sequence scans. Moreover, a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan. DB1/ 147040794.2 7 Attorney Ref. No.: 115834-5037-WO [0038] Another aspect of the present disclosure is directed to providing computer system. The computer system includes one or more processors, a controller, and at least one program is non-transiently stored in the controller and executable by the controller. The at least one program cause the controller to perform a method. In some embodiments, the method includes obtaining, in electronic form, a training data set. The training data set includes a first plurality of graphical data. The first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data. Moreover, the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data. In addition, in some embodiments, the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data. The performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances. Moreover, the method includes receiving a request to evaluation a second one or more graphical data at a second computer system. Moreover, the second one or more graphical data is defined by a second plurality of characteristics. The second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data. Furthermore, the second plurality of characteristics includes a desired threshold performance. The method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. Furthermore, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. Furthermore, the method includes matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data. Additionally, the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream and one or more instructions to terminate DB1/ 147040794.2 8 Attorney Ref. No.: 115834-5037-WO decoding of the encoded byte stream at the second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, thereby optimizing decoding of graphical. [0039] Yet another aspect of the present disclosure is directed to providing a non-transitory computer readable storage medium storing one or more programs. The one or more programs include instructions, which when executed by a computer system, cause the computer system to perform a method. In some embodiments, the method includes obtaining, in electronic form, a training data set. The training data set includes a first plurality of graphical data. The first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data. Moreover, the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data. In addition, in some embodiments, the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data. The performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances. Moreover, the method includes receiving a request to evaluation a second one or more graphical data at a second computer system. Moreover, the second one or more graphical data is defined by a second plurality of characteristics. The second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data. Furthermore, the second plurality of characteristics includes a desired threshold performance. The method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. Furthermore, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. Furthermore, the method includes matching the desired DB1/ 147040794.2 9 Attorney Ref. No.: 115834-5037-WO threshold performance to a first training resolution in the plurality of training resolutions in the performance data. Additionally, the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream at the second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, thereby optimizing decoding of graphical. [0040] Yet another aspect of the present disclosure is directed to providing a method for optimizing encoding of graphical data. The method includes using a first computer system. The first computer system includes a processor and memory. In some embodiments, the method includes obtaining, in electronic form, a training data set. The training data set includes a first plurality of graphical data. The first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data. Moreover, the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data. In addition, in some embodiments, the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data. The performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances. Moreover, the method includes receiving a request to evaluation a second one or more graphical data at a second computer system. Moreover, the second one or more graphical data is defined by a second plurality of characteristics. The second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data. Furthermore, the second plurality of characteristics includes a desired threshold performance. The method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. Furthermore, each sequence scan in the sequence of the plurality of sequence scans has a DB1/ 147040794.2 10 Attorney Ref. No.: 115834-5037-WO corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. [0041] In some embodiments, the first plurality of graphical data includes one or more images taken of a first biological sample. [0042] In some embodiments, the second one or more graphical data includes one or more images taken of a second biological sample different than the first biological sample. [0043] In some embodiments, the first plurality of graphical data includes as 2D graphical data, 3D graphical data, 4D graphical data, or a combination thereof. [0044] In some embodiments, the second one or more graphical data includes 3D graphical data. Moreover, each scan in the plurality of sequence scans of the 3D graphical data includes a respective 2D layer of the 3D graphical data. [0045] In some embodiments, the first plurality of graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. [0046] In some embodiments, the first plurality of graphical data includes 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images. [0047] In some embodiments, the second one or more graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non- fungible token assets, or a combination thereof. [0048] In some embodiments, the sampling resolution of the first plurality of graphical data includes the minimum resolution associated with the first plurality of graphical data. [0049] In some embodiments, the first modality includes a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof. DB1/ 147040794.2 11 Attorney Ref. No.: 115834-5037-WO [0050] In some embodiments, the first codec includes a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof. [0051] In some embodiments, the first codec is selected by the first computer system. [0052] In some embodiments, the plurality of sequence scans includes between 2 scans and 100 scans. [0053] In some embodiments, the plurality of sequence scans includes N scans, in which In some embodiments, M is a number of digital assets associated with
Figure imgf000014_0001
the second one or more graphical data. In some embodiments, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data. Moreover, in some embodiments, ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. [0054] In some embodiments, the plurality of sequence scans includes N scans, in which ^^^^ = ^^^^ ^^^^ 2 − 1, M is a number of digital assets associated with the second one or more
Figure imgf000014_0002
graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. [0055] In some embodiments, the first plurality of characteristics further includes one or more ground truth labels with respect to the first modality for the first plurality of graphical data. [0056] In some embodiments, the using the first plurality of graphical data further includes training the feature extraction model against the one or more ground truth labels. [0057] In some embodiments, the second plurality of characteristics includes a sampling resolution of the second one or more graphical data. Moreover, in some embodiments, the plurality of sequence scans includes N scans that is based, at least in part, on the sampling resolution of the second one or more graphical data. [0058] In some embodiments, the second plurality of characteristics further includes a respective capacity of the communication network. [0059] In some embodiments, the second plurality of characteristics further includes a respective capacity of the first computer system. DB1/ 147040794.2 12 Attorney Ref. No.: 115834-5037-WO [0060] In some embodiments, the feature extraction model includes a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof. [0061] In some embodiments, the feature extraction model incudes a neutral network model, a support machine mode, a Naïve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph-based model, or a combination thereof. [0062] In some embodiments, the desired threshold performance includes a threshold inter- image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation. [0063] In some embodiments, the encoded byte stream is associated with a first file size of the final terminal scan in the plurality of sequence scans. Moreover, a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan. [0064] Another aspect of the present disclosure is directed to providing computer system. The computer system includes one or more processors, a controller, and at least one program is non-transiently stored in the controller and executable by the controller. The at least one program cause the controller to perform a method. In some embodiments, the method includes obtaining, in electronic form, a training data set. The training data set includes a first plurality of graphical data. The first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data. Moreover, the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data. In addition, in some embodiments, the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data. The performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances. Moreover, the method includes receiving a request to evaluation a second one or more graphical data at a second computer system. Moreover, the second one or more graphical data is defined by a second plurality of characteristics. The second plurality of characteristics includes the first modality associated DB1/ 147040794.2 13 Attorney Ref. No.: 115834-5037-WO with a capture of the second one or more graphical data. Furthermore, the second plurality of characteristics includes a desired threshold performance. The method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. Furthermore, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. [0065] Yet another aspect of the present disclosure is directed to providing a non-transitory computer readable storage medium storing one or more programs. The one or more programs include instructions, which when executed by a computer system, cause the computer system to perform a method. In some embodiments, the method includes obtaining, in electronic form, a training data set. The training data set includes a first plurality of graphical data. The first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data. Moreover, the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data. In addition, in some embodiments, the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data. The performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances. Moreover, the method includes receiving a request to evaluation a second one or more graphical data at a second computer system. Moreover, the second one or more graphical data is defined by a second plurality of characteristics. The second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data. Furthermore, the second plurality of characteristics includes a desired threshold performance. The method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this DB1/ 147040794.2 14 Attorney Ref. No.: 115834-5037-WO encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. Furthermore, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. [0066] Yet another aspect of the present disclosure is directed to providing a method of optimizing communication of graphical data. The method includes using a first computer system. The first computer system includes a processor and memory. In some embodiments, the method includes obtaining, in electronic form, a training data set. The training data set includes a first plurality of graphical data. The first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data. Moreover, the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data. In addition, in some embodiments, the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data. The performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances. Moreover, the method includes receiving a request to evaluation a second one or more graphical data at a second computer system. Moreover, the second one or more graphical data is defined by a second plurality of characteristics. The second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data. Furthermore, the second plurality of characteristics includes a desired threshold performance. The method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. DB1/ 147040794.2 15 Attorney Ref. No.: 115834-5037-WO Furthermore, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. Furthermore, the method includes matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data. Additionally, the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream. Moreover, the method includes terminating, in accordance with a determination that decoding of the encoded byte stream at the second computer system satisfies a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, the communicating of the encoded byte stream, thereby optimizing communication of graphical data. [0067] In some embodiments, the first plurality of graphical data includes one or more images taken of a first biological sample. [0068] In some embodiments, the second one or more graphical data includes one or more images taken of a second biological sample different than the first biological sample. [0069] In some embodiments, the first plurality of graphical data includes as 2D graphical data, 3D graphical data, 4D graphical data, or a combination thereof. [0070] In some embodiments, the second one or more graphical data includes 3D graphical data. Moreover, each scan in the plurality of sequence scans of the 3D graphical data includes a respective 2D layer of the 3D graphical data. [0071] In some embodiments, the first plurality of graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. [0072] In some embodiments, the first plurality of graphical data includes 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images. [0073] In some embodiments, the second one or more graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non- fungible token assets, or a combination thereof. DB1/ 147040794.2 16 Attorney Ref. No.: 115834-5037-WO [0074] In some embodiments, the sampling resolution of the first plurality of graphical data includes the minimum resolution associated with the first plurality of graphical data. [0075] In some embodiments, the first modality includes a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof. [0076] In some embodiments, the first codec includes a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof. [0077] In some embodiments, the first codec is selected by the first computer system. [0078] In some embodiments, the plurality of sequence scans includes between 2 scans and 100 scans. [0079] In some embodiments, the plurality of sequence scans includes N scans, in which In some embodiments, M is a number of digital assets associated with
Figure imgf000019_0001
the second one or more graphical data. In some embodiments, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data. Moreover, in some embodiments, ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. [0080] In some embodiments, the plurality of sequence scans includes N scans, in which ^^^^ = ^^^^ ^^^^ ^^^^ ^^^^ ∑ ^^^^=1 ^^^^ ^^^^ 2� ^^^^ ^^^^ ^^^^ ^^^^ ^^^^� − 1, M is a number of digital assets associated with the second one or more graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. [0081] In some embodiments, the first plurality of characteristics further includes one or more ground truth labels with respect to the first modality for the first plurality of graphical data. [0082] In some embodiments, the using the first plurality of graphical data further includes training the feature extraction model against the one or more ground truth labels. [0083] In some embodiments, the second plurality of characteristics includes a sampling resolution of the second one or more graphical data. Moreover, in some embodiments, the DB1/ 147040794.2 17 Attorney Ref. No.: 115834-5037-WO plurality of sequence scans includes N scans that is based, at least in part, on the sampling resolution of the second one or more graphical data. [0084] In some embodiments, the second plurality of characteristics further includes a respective capacity of the communication network. [0085] In some embodiments, the second plurality of characteristics further includes a respective capacity of the first computer system. [0086] In some embodiments, the feature extraction model includes a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof. [0087] In some embodiments, the feature extraction model incudes a neutral network model, a support machine mode, a Naïve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph-based model, or a combination thereof. [0088] In some embodiments, the first sequence scan includes less than 5% of the encoded byte stream. [0089] In some embodiments, the first sequence scan includes less than 5% of the final terminal scan. [0090] In some embodiments, the desired threshold performance includes a threshold inter- image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation. [0091] In some embodiments, the encoded byte stream is associated with a first file size of the final terminal scan in the plurality of sequence scans. Moreover, a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan. [0092] Another aspect of the present disclosure is directed to providing computer system. The computer system includes one or more processors, a controller, and at least one program is non-transiently stored in the controller and executable by the controller. The at least one program cause the controller to perform a method. In some embodiments, the method includes obtaining, in electronic form, a training data set. The training data set includes a first plurality of graphical data. The first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data. Moreover, the first plurality of characteristics includes a first modality DB1/ 147040794.2 18 Attorney Ref. No.: 115834-5037-WO associated with a capture of the first plurality of graphical data. In addition, in some embodiments, the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data. The performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances. Moreover, the method includes receiving a request to evaluation a second one or more graphical data at a second computer system. Moreover, the second one or more graphical data is defined by a second plurality of characteristics. The second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data. Furthermore, the second plurality of characteristics includes a desired threshold performance. The method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. Furthermore, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. Furthermore, the method includes matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data. Additionally, the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream. Moreover, the method includes terminating, in accordance with a determination that decoding of the encoded byte stream at the second computer system satisfies a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, the communicating of the encoded byte stream, thereby optimizing communication of graphical data. [0093] Yet another aspect of the present disclosure is directed to providing a non-transitory computer readable storage medium storing one or more programs. The one or more programs include instructions, which when executed by a computer system, cause the DB1/ 147040794.2 19 Attorney Ref. No.: 115834-5037-WO computer system to perform a method. In some embodiments, the method includes obtaining, in electronic form, a training data set. The training data set includes a first plurality of graphical data. The first plurality of graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first plurality of graphical data. Moreover, the first plurality of characteristics includes a first modality associated with a capture of the first plurality of graphical data. In addition, in some embodiments, the method includes using the first plurality of graphical data to train a feature extraction model. From this using of the first plurality of graphical data, the method obtains a trained feature extraction model and performance data. The performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances. Moreover, the method includes receiving a request to evaluation a second one or more graphical data at a second computer system. Moreover, the second one or more graphical data is defined by a second plurality of characteristics. The second plurality of characteristics includes the first modality associated with a capture of the second one or more graphical data. Furthermore, the second plurality of characteristics includes a desired threshold performance. The method further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data. From this encoding, the method forms an encoded byte stream that includes a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. Furthermore, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. Furthermore, the method includes matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data. Additionally, the method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream. Moreover, the method includes terminating, in accordance with a determination that decoding of the encoded byte stream at the second computer system satisfies a first sequence scan in the plurality of sequence scans that matches or exceeds the DB1/ 147040794.2 20 Attorney Ref. No.: 115834-5037-WO first training resolution, the communicating of the encoded byte stream, thereby optimizing communication of graphical data. [0094] Yet another aspect of the present disclosure is directed to providing a method for optimizing decoding of graphical data. The method includes obtaining, in electronic form, a first one or more graphical data. The first one or more graphical data is defined by a first plurality of characteristics that includes a sampling resolution of the first one or more of graphical data. The first one or more graphical data is further defined a first modality associated with a capture of the first one or more of graphical data. The method includes forming a byte stream map for the first one or more graphical data. In some embodiments, the byte stream map includes a plurality of sequence scans through the sampling resolution. In some embodiments, each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans. Moreover, in some embodiments, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. Additionally, the method includes receiving, from a second computer system, a request to evaluate the first one or more graphical data at a first resolution. The first resolution is selected in accordance with a desired threshold performance from performance data. The performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Each respective training resolution is associated with a corresponding threshold performance for a trained feature extraction model in a plurality of threshold performances. Each threshold performance in the plurality of threshold performances associated with the first modality associated with a capture of a second plurality of graphical data. The method further includes matching the first resolution to a first sequence scan in the plurality of sequence scans in accordance with the byte stream map. Furthermore, the method includes encoding, in accordance with a corresponding encoder of a first codec, the first one or more graphical data, thereby forming an encoded byte stream that includes the plurality of sequence scans through at least the first resolution. The method includes communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream at the second computer system when the DB1/ 147040794.2 21 Attorney Ref. No.: 115834-5037-WO decoding decodes the first sequence scan in the plurality of sequence scans, thereby optimizing decoding of graphical data. [0095] In some embodiments, the first one or more of graphical data includes one or more images taken of a first biological sample. [0096] In some embodiments, the first one or more graphical data includes one or more images taken of a second biological sample different than the first biological sample. [0097] In some embodiments, the sampling resolution of the first one or more graphical data includes the minimum resolution associated with the first one or more graphical data. [0098] In some embodiments, the first modality includes a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof. [0099] In some embodiments, the first plurality of characteristics further includes one or more ground truth labels with respect to the first modality for the first one or more of graphical data. [00100] In some embodiments, the first plurality of characteristics further includes a respective capacity of the communication network. [00101] In some embodiments, the first plurality of characteristics further includes a respective capacity of the second computer system. [00102] In some embodiments, the first one or more graphical data includes 3D graphical data and each scan in the plurality of sequence scans of the 3D graphical data includes a respective 2D layer of the 3D graphical data [00103] In some embodiments, the first one or more graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non- fungible token assets, or a combination thereof. [00104] In some embodiments, the plurality of sequence scans includes between 2 scans and 100 scans. [00105] In some embodiments, the plurality of sequence scans includes N scans, in which In some embodiments, M is a number of digital assets associated with
Figure imgf000024_0001
DB1/ 147040794.2 22 Attorney Ref. No.: 115834-5037-WO the second one or more graphical data. In some embodiments, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data. Moreover, in some embodiments, ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. [00106] In some embodiments, the plurality of sequence scans includes N scans, in which ^^^^ − 1, M is a number of digital assets associated with the second one or
Figure imgf000025_0001
more graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. [00107] In some embodiments, the first plurality of characteristics includes a sampling resolution of the first one or more graphical data, in which the plurality of sequence scans includes N scans based, at least in part, on the sampling resolution of the first one or more graphical data. [00108] In some embodiments, the second plurality of graphical data includes two- dimensional (2D) graphical data, three-dimensional (3D) graphical data, four-dimensional (4D) data, or a combination thereof. [00109] In some embodiments, the second plurality of graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non- fungible token assets, or a combination thereof. [00110] In some embodiments, the second plurality of graphical data includes 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images. [00111] In some embodiments, prior to the receiving C), the method further includes training the feature extraction model against the one or more ground truth labels. [00112] In some embodiments, the desired threshold performance includes a threshold inter- image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation. [00113] In some embodiments, the feature extraction model includes a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof. DB1/ 147040794.2 23 Attorney Ref. No.: 115834-5037-WO [00114] In some embodiments, the feature extraction model includes a neutral network model, a support machine mode, a Naïve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph-based model, or a combination thereof. [00115] In some embodiments, the first codec includes a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof. [00116] In some embodiments, the first codec is selected by the second computer system. [00117] In some embodiments, the first sequence scan includes less than 5% of the encoded byte stream. [00118] In some embodiments, the first sequence scan includes less than 5% of a final terminal scan. [00119] In some embodiments, the encoded byte stream is associated with a first file size of the final terminal scan in the plurality of sequence scans, and a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan. [00120] The systems, methods, devices, and non-transitory computer readable storage medium of the present invention have other features and advantages that will be apparent from, or are set forth in more detail in, the accompanying drawings, which are incorporated herein, and the following Detailed Description, which together serve to explain certain principles of exemplary embodiments of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS [00121] Figure 1 illustrates an exemplary system topology including a distributed computer system including a graphical data system and one or more client devices, in accordance with an exemplary embodiment of the present disclosure; [00122] Figures 2A and 2B collectively illustrates a graphical data system for generating an encoded byte stream, matching a desired threshold performance associated with a request to evaluate graphical data to a first training resolution, communicating the encoded byte stream, DB1/ 147040794.2 24 Attorney Ref. No.: 115834-5037-WO terminating communication of the encoded byte stream, or a combination thereof, in accordance with an embodiment of the present disclosure; [00123] Figure 3 illustrates a client device, in accordance with an embodiment of the present disclosure; [00124] Figures 4A, 4B, 4C, 4D, and 4E collectively illustrate exemplary methods for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof, in which optional embodiments are indicated by dashed boxes, in accordance with some embodiments of the present disclosure; [00125] Figures 5A, 5B, 5C, 5D, and 5E collectively illustrate exemplary methods for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof, in which optional embodiments are indicated by dashed boxes, in accordance with some embodiments of the present disclosure; [00126] Figure 6 illustrates a chart depicting an overview of an architecture for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof, in accordance with some embodiments of the present disclosure; [00127] Figure 7 illustrates a chart depicting a comparison of graphical data and model metrics for two-dimensional (2D) graphical data applied to a classification model using a first portable network graphics (PNG) codec and a second high-throughput joint photographic expert group (JPEG) 2000 (HTJ2K) codec for a first plurality of graphical data (e.g., a held- out NIH test set), in accordance with some embodiments of the present disclosure; [00128] Figure 8 illustrates a chart depicting a comparison of graphical data and model metrics for 2D graphical data applied to a classification model using a first PNG codec and a second HTJ2K codec for a second one or more graphical data (e.g., a MIMIC external test set), in accordance with some embodiments of the present disclosure; [00129] Figure 9 illustrates a chart depicting a comparison of graphical data and performance data for three-dimensional (3D) graphical data applied to a segmentation feature extraction model for a first liver and spleen modality segmentation using a first neuroimaging informatics technology initiative (NifTI) codec and a second HTJ2K version of a third plurality of graphical data (e.g., a held-out Medical Segmentation Decathlon (MSD) liver and spleen test data set), in accordance with some embodiments of the present disclosure; DB1/ 147040794.2 25 Attorney Ref. No.: 115834-5037-WO [00130] Figure 10 illustrates a chart depicting a comparison of graphical data and model metrics for 3D graphical data applied to a segmentation model for liver and spleen segmentation using a first NifTI codec and a second HTJ2K version of a second plurality of graphical data (e.g., the Multi-Atlas Labelling Beyond the Cranial Vault (BTCV) external test data set), in accordance with some embodiments of the present disclosure; [00131] Figure 11 illustrates a representation of a plurality of sequence scans of an encoded byte stream for a second one or more graphical data, in accordance with some embodiments of the present disclosure; [00132] Figure 12 illustrates a block diagram of a HTJ2K codec including an encoder and a decoder for communicating an encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream, in accordance with some embodiments of the present disclosure; [00133] Figure 13A illustrates overview of a conventional architecture in which the host encodes imaging data using non-progressive codecs (e.g., JPEG), in which, for inference, the host must stream the complete byte stream before the medical image is processed and characterized by an AI system; [00134] Figure 13B overview of an architecture for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof, in accordance with some embodiments of the present disclosure; [00135] Figure 14 illustrates a representation of progressive encoding and subsequent decoding at different decomposition levels to access sub-resolution images by selecting different subsets of the byte stream for chest x-ray, in accordance with some embodiments of the present disclosure; [00136] Figure 15A illustrate a representation of encoding of a chest x-ray using JPEG (sequential) and HTJ2K (progressive) codecs, followed by decoding of partial byte streams from both formats, in accordance with some embodiments of the present disclosure; [00137] Figure 15B illustrate a representation of an encoded byte stream utilized to reconstruct graphical data from a respective sequence scan having a lower resolution than the original graphical data, in accordance with some embodiments of the present disclosure; [00138] Figure 16 illustrates an overview of architecture for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a DB1/ 147040794.2 26 Attorney Ref. No.: 115834-5037-WO combination thereof, in which at a graphical data system, a collection of graphical data (e.g., uncompressed DICOM chest x-rays) are progressively encoded into encoded byte streams using an encoded (e.g., HTJ2K), at host-side and client-side stream optimizers interact with each other to stream partial encoded byte streams from the graphical data system to an client device, and at the client device, the partial encoded byte streams are progressively decoded to reconstruct the graphical data at the optimal resolution for evaluation by an AI model in accordance with a matched sequence scan of the optimal resolution, in accordance with some embodiments of the present disclosure; [00139] Figures 17A, 17B, and 17C collectively depict charts of mean AUROC scores for 2D classification of abnormalities using chest x-ray graphical data on the original graphical data and across each HTJ2K scan, including an optimal resolution (hatched) for an Intelligent Streaming framework (“ISLE”), using the original graphical data such as the internal NIH test set (Figure 17A), external CheXpert dataset (Figure 17B), and external MIMIC dataset (Figure 17C), in which performance metrics were compared using one-sided paired t-tests, the comparison between the original dataset and the optimal resolution is annotated, and for all other comparisons, only statistically significant comparisons are annotated (ns: P > 0.05, *: P ≤ 0.05, **: P ≤ 0.01, ***: P ≤ 0.001), in accordance with some embodiments of the present disclosure; [00140] Figures 18A and 18B collectively depict charts of mean Dice scores for 3D segmentation of the liver (Figure 18A) and (Figure 18B) spleen organs using abdomen CT scans on the original graphical data and across each HTJ2K scans, including the ISLE’s optimal resolution (hatched), using the internal MSD test set and external BTCV dataset, in which performance metrics were compared using one-sided Wilcoxon rank-sum tests, the comparison between the original dataset and the optimal resolution is annotated, and for all other comparisons, only statistically significant comparisons are annotated (ns: P > 0.05, *: P ≤ 0.05, **: P ≤ 0.01, ***: P ≤ 0.001), in accordance with some embodiments of the present disclosure; [00141] Figure 19 illustrates examples of progressively encoded sequence scans of chest x- rays and abdomen CT graphical data from the NIH and MSD Liver dataset respectively across all scan levels, in accordance with some embodiments of the present disclosure; [00142] Figure 20 illustrates progressively encoded volumes from the MSD Liver dataset decoded across the three decomposition levels to reconstruct volumes with axial sub- DB1/ 147040794.2 27 Attorney Ref. No.: 115834-5037-WO resolution 64x64, 128x128, and 256x256 as well as the lossless reconstruction at resolution 512x512, in accordance with some embodiments of the present disclosure; and [00143] Figure 21 illustrates exemplary logical functions that are used implemented in various embodiments of the present disclosure. [00144] It should be understood that the appended drawings are not necessarily to scale, presenting a somewhat simplified representation of various features illustrative of the basic principles of the invention. The specific design features of the present invention as disclosed herein, including, for example, specific dimensions, orientations, locations, and shapes will be determined in part by the particular intended application and use environment. [00145] In the figures, reference numbers refer to the same or equivalent parts of the present invention throughout the several figures of the drawing. DETAILED DESCRIPTION [00146] The present disclosure provides systems and methods for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof are provided. A training data set including first graphical data defined by a sampling resolution and a first modality are obtained. The first graphical data (e.g., first plurality of graphical data) is used to train a feature extraction model. From this training, a trained feature extraction model and performance data are obtained. The performance data includes a plurality of training resolutions. Each respective training resolution in the plurality of training resolutions is less than the sampling resolution of the first graphical data. Moreover, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances. Accordingly, in some embodiments, the plurality of training resolutions together with the plurality of threshold performances provide for a standard of comparison for future applications with a respective feature extraction model. A request to evaluate second graphical data (e.g., second one or more graphical data) is received. The second graphical data is defined by the first modality and a desired threshold performance. In some embodiments, the second graphical data is distinct from the first graphical data. The second graphical data is encoded forming an encoded byte stream including a plurality of sequence scans through the sampling resolution. Each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of DB1/ 147040794.2 28 Attorney Ref. No.: 115834-5037-WO respective sequence scans in the plurality of sequence scans. Moreover, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. Accordingly, in some embodiments, the corresponding progressive resolution allows for each respective sequence scan to have a corresponding unique resolution. In some embodiments, the desired threshold performance is matched to a first training resolution in the plurality of training resolutions in the performance data. Furthermore, in some embodiments, the encoded byte stream is communicated to the second computer system, via a communication network, in accordance with the first codec. In some embodiments, the communicating the encoded byte stream includes one or more instructions to terminate decoding of the encoded byte stream at a second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution. Still further, in some embodiments, the one or more instructions is configured to terminate communication of the encoded byte stream at a second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution. Accordingly, the systems and methods of the present disclosure allow for optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof [00147] Reference will now be made in detail to embodiments, examples of which are illustrated in the accompanying drawings. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the present disclosure. However, it will be apparent to one of ordinary skill in the art that the present disclosure may be practiced without these specific details. In other instances, well-known methods, procedures, and components have not been described in detail so as not to unnecessarily obscure aspects of the embodiments. [00148] It will also be understood that, although the terms first, second, etc. may be used herein to describe various elements, these elements should not be limited by these terms. These terms are only used to distinguish one element from another. For instance, a first subject could be termed a second subject, and, similarly, a second subject could be termed a first subject, without departing from the scope of the present disclosure. The first subject and the second subject are both subjects, but they are not the same subject. DB1/ 147040794.2 29 Attorney Ref. No.: 115834-5037-WO [00149] The terminology used in the present disclosure is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the description of the invention and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that the term “and/or” as used herein refers to and encompasses any and all possible combinations of one or more of the associated listed items. It will be further understood that the terms “comprises” and/or “comprising,” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. [00150] The foregoing description included example systems, methods, techniques, instruction sequences, and computing machine program products that embody illustrative implementations. For purposes of explanation, numerous specific details are set forth in order to provide an understanding of various implementations of the inventive subject matter. It will be evident, however, to those skilled in the art that implementations of the inventive subject matter may be practiced without these specific details. In general, well-known instruction instances, protocols, structures, and techniques have not been shown in detail. [00151] The foregoing description, for purpose of explanation, has been described with reference to specific implementations. However, the illustrative discussions below are not intended to be exhaustive or to limit the implementations to the precise forms disclosed. Many modifications and variations are possible in view of the above teachings. The implementations are chosen and described in order to best explain the principles and their practical applications, to thereby enable others skilled in the art to best utilize the implementations and various implementations with various modifications as are suited to the particular use contemplated. [00152] In the interest of clarity, not all of the routine features of the implementations described herein are shown and described. It will be appreciated that, in the development of any such actual implementation, numerous implementation-specific decisions are made in order to achieve the designer’s specific goals, such as compliance with use case- and business-related constraints, and that these specific goals will vary from one implementation to another and from one designer to another. Moreover, it will be appreciated that such a design effort might be complex and time-consuming, but nevertheless be a routine DB1/ 147040794.2 30 Attorney Ref. No.: 115834-5037-WO undertaking of engineering for those of ordering skill in the art having the benefit of the present disclosure. [00153] As used herein, the term “if” may be construed to mean “when” or “upon” or “in response to determining” or “in response to detecting,” depending on the context. Similarly, the phrase “if it is determined” or “if [a stated condition or event] is detected” may be construed to mean “upon determining” or “in response to determining” or “upon detecting [the stated condition or event]” or “in response to detecting [the stated condition or event],” depending on the context. [00154] Furthermore, as used herein, the term “about” or “approximately” can mean within an acceptable error range for the particular value as determined by one of ordinary skill in the art, which can depend in part on how the value is measured or determined, e.g., the limitations of the measurement system. For example, “about” can mean within 1 or more than 1 standard deviation, per the practice in the art. “About” can mean a range of ± 20%, ± 10%, ± 5%, or ± 1% of a given value. Where particular values are described in the application and claims, unless otherwise stated, the term “about” means within an acceptable error range for the particular value. The term “about” can have the meaning as commonly understood by one of ordinary skill in the art. The term “about” can refer to ± 10%. The term “about” can refer to ± 5%. [00155] As used herein, the term “lossless,” means a compression ratio of 0% or about 0%. [00156] As used herein, the term “lossy,” means a compression ratio of less or equal to 5%. [00157] As used herein, the term “scan,” “sequence scan” and “decomposition” are used interchangeably herein unless expressly stated otherwise. [00158] Moreover, as used herein, the term “AxB,” denotes a resolution of two-dimensional graphical data in which “A” is a number of pixels in a horizontal direction and “B” is a number of pixels in a vertical direction. [00159] Moreover, as used herein, the term “AxBxC,” denotes a resolution of three- dimensional graphical data in which “A” is a number of pixels in a horizontal direction, “B” is a number of pixels in a vertical direction, and “C” is a number of pixels in a depth direction. [00160] As used interchangeably herein, the term “classifier” or “model” refers to a machine learning model or algorithm. DB1/ 147040794.2 31 Attorney Ref. No.: 115834-5037-WO [00161] In some embodiments, a model includes an unsupervised learning algorithm. One example of an unsupervised learning algorithm is cluster analysis. In some embodiments, a model includes supervised machine learning. Nonlimiting examples of supervised learning algorithms include, but are not limited to, logistic regression, neural networks, support vector machines, Naive Bayes algorithms, nearest neighbor algorithms, random forest algorithms, decision tree algorithms, boosted trees algorithms, multinomial logistic regression algorithms, linear models, linear regression, Gradient Boosting, mixture models, hidden Markov models, Gaussian NB algorithms, linear discriminant analysis, or any combinations thereof. In some embodiments, a model is a multinomial classifier algorithm. In some embodiments, a model is a 2-stage stochastic gradient descent (SGD) model. In some embodiments, a model is a deep neural network (e.g., a deep-and-wide sample-level model). [00162] Neural networks. In some embodiments, the model is a neural network (e.g., a convolutional neural network and/or a residual neural network). Neural network algorithms, also known as artificial neural networks (ANNs), include convolutional and/or residual neural network algorithms (deep learning algorithms). In some embodiments, neural networks are machine learning algorithms that are trained to map an input dataset to an output dataset, where the neural network includes an interconnected group of nodes organized into multiple layers of nodes. For example, in some embodiments, the neural network architecture includes at least an input layer, one or more hidden layers, and an output layer. In some embodiments, the neural network includes any total number of layers, and any number of hidden layers, where the hidden layers function as trainable feature extractors that allow mapping of a set of input data to an output value or set of output values. In some embodiments, a deep learning algorithm is a neural network including a plurality of hidden layers, e.g., two or more hidden layers. In some instances, each layer of the neural network includes a number of nodes (or “neurons”). In some embodiments, a node receives input that comes either directly from the input data or the output of nodes in previous layers, and performs a specific operation, e.g., a summation operation. In some embodiments, a connection from an input to a node is associated with a parameter (e.g., a weight and/or weighting factor). In some embodiments, the node sums up the products of all pairs of inputs, xi, and their associated parameters. In some embodiments, the weighted sum is offset with a bias, b. In some embodiments, the output of a node or neuron is gated using a threshold or activation function, f, which, in some instances, is a linear or non-linear function. In some embodiments, the activation function is, for example, a rectified linear unit (ReLU) DB1/ 147040794.2 32 Attorney Ref. No.: 115834-5037-WO activation function, a Leaky ReLU activation function, or other function such as a saturating hyperbolic tangent, identity, binary step, logistic, arcTan, softsign, parametric rectified linear unit, exponential linear unit, softPlus, bent identity, softExponential, Sinusoid, Sine, Gaussian, or sigmoid function, or any combination thereof. [00163] In some implementations, the weighting factors, bias values, and threshold values, or other computational parameters of the neural network, are “taught” or “learned” in a training phase using one or more sets of training data. For example, in some implementations, the parameters are trained using the input data from a training dataset and a gradient descent or backward propagation method so that the output value(s) that the ANN computes are consistent with the examples included in the training dataset. In some embodiments, the parameters are obtained from a back propagation neural network training process. [00164] Any of a variety of neural networks are suitable for use in accordance with the present disclosure. Examples include, but are not limited to, graph neural networks, feedforward neural networks, radial basis function networks, recurrent neural networks, residual neural networks, convolutional neural networks, residual convolutional neural networks, and the like, or any combination thereof. In some embodiments, the machine learning makes use of a pre-trained and/or transfer-learned ANN or deep learning architecture. In some implementations, convolutional and/or residual neural networks are used, in accordance with the present disclosure. [00165] For instance, a deep neural network model includes an input layer, a plurality of individually parameterized (e.g., weighted) convolutional layers, and an output scorer. The parameters (e.g., weights) of each of the convolutional layers as well as the input layer contribute to the plurality of parameters (e.g., weights) associated with the deep neural network model. In some embodiments, at least 50 parameters, at least 100 parameters, at least 1,000 parameters, at least 2,000 parameters or at least 5,000 parameters are associated with the deep neural network model. As such, deep neural network models require a computer to be used because they cannot be mentally solved. In other words, given an input to the model, the model output needs to be determined using a computer rather than mentally in such embodiments. See, for example, Krizhevsky et al., 2012, “Imagenet classification with deep convolutional neural networks,” in Advances in Neural Information Processing Systems 2, Pereira, Burges, Bottou, Weinberger, eds., pp.1097-1105, Curran Associates, Inc.; Zeiler, 2012 “ADADELTA: an adaptive learning rate method,” CoRR, vol. abs/1212.5701; and Rumelhart et al., 1988, “Neurocomputing: Foundations of research,” ch. Learning DB1/ 147040794.2 33 Attorney Ref. No.: 115834-5037-WO Representations by Back-propagating Errors, pp.696-699, Cambridge, MA, USA: MIT Press, each of which is hereby incorporated by reference. [00166] Neural network algorithms, including convolutional neural network algorithms, suitable for use as models are disclosed in, for example, Vincent et al., 2010, “Stacked denoising autoencoders: Learning useful representations in a deep network with a local denoising criterion,” J Mach Learn Res 11, pp.3371-3408; Larochelle et al., 2009, “Exploring strategies for training deep neural networks,””dec J Mach Learn Res 10, pp.1-40; and Hassoun, 1995, Fundamentals of Artificial Neural Networks, Massachusetts Institute of Technology, each of which is hereby incorporated by reference. Additional example neural networks suitable for use as models are disclosed in Duda et al., 2001, Pattern Classification, Second Edition, John Wiley & Sons, Inc., New York; and Hastie et al., 2001, The Elements of Statistical Learning, Springer-Verlag, New York, each of which is hereby incorporated by reference in its entirety. Additional example neural networks suitable for use as models are also described in Draghici, 2003, Data Analysis Tools for DNA Microarrays, Chapman & Hall/CRC; and Mount, 2001, Bioinformatics: sequence and genome analysis, Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York, each of which is hereby incorporated by reference in its entirety. [00167] Ensembles of models and boosting. In some embodiments, an ensemble (two or more) of models is used. In some embodiments, a boosting technique such as AdaBoost is used in conjunction with many other types of learning algorithms to improve the performance of the model. In this approach, the output of any of the models disclosed herein, or their equivalents, is combined into a weighted sum that represents the final output of the boosted model. In some embodiments, the plurality of outputs from the models is combined using any measure of central tendency known in the art, including but not limited to a mean, median, mode, a weighted mean, weighted median, weighted mode, etc. In some embodiments, the plurality of outputs is combined using a voting method. In some embodiments, a respective model in the ensemble of models is weighted or unweighted. [00168] As used herein, the term “parameter” refers to any coefficient or, similarly, any value of an internal or external element (e.g., a weight and/or a hyperparameter) in an algorithm, model, regressor, and/or classifier that can affect (e.g., modify, tailor, and/or adjust) one or more inputs, outputs, and/or functions in the algorithm, model, regressor and/or classifier. For example, in some embodiments, a parameter refers to any coefficient, weight, and/or hyperparameter that can be used to control, modify, tailor, and/or adjust the behavior, DB1/ 147040794.2 34 Attorney Ref. No.: 115834-5037-WO learning, and/or performance of an algorithm, model, regressor, and/or classifier. In some instances, a parameter is used to increase or decrease the influence of an input (e.g., a feature) to an algorithm, model, regressor, and/or classifier. As a nonlimiting example, in some embodiments, a parameter is used to increase or decrease the influence of a node (e.g., of a neural network), where the node includes one or more activation functions. Assignment of parameters to specific inputs, outputs, and/or functions is not limited to any one paradigm for a given algorithm, model, regressor, and/or classifier but can be used in any suitable algorithm, model, regressor, and/or classifier architecture for a desired performance. In some embodiments, a parameter has a fixed value. In some embodiments, a value of a parameter is manually and/or automatically adjustable. In some embodiments, a value of a parameter is modified by a validation and/or training process for an algorithm, model, regressor, and/or classifier (e.g., by error minimization and/or backpropagation methods). In some embodiments, an algorithm, model, regressor, and/or classifier of the present disclosure includes a plurality of parameters. In some embodiments, the plurality of parameters is n parameters, where: n ≥ 2; n ≥ 5; n ≥ 10; n ≥ 25; n ≥ 40; n ≥ 50; n ≥ 75; n ≥ 100; n ≥ 125; n ≥ 150; n ≥ 200; n ≥ 225; n ≥ 250; n ≥ 350; n ≥ 500; n ≥ 600; n ≥ 750; n ≥ 1,000; n ≥ 2,000; n ≥ 4,000; n ≥ 5,000; n ≥ 7,500; n ≥ 10,000; n ≥ 20,000; n ≥ 40,000; n ≥ 75,000; n ≥ 100,000; n ≥ 200,000; n ≥ 500,000, n ≥ 1 x 106, n ≥ 5 x 106, or n ≥ 1 x 107. As such, the algorithms, models, regressors, and/or classifiers of the present disclosure cannot be mentally performed. In some embodiments n is between 10,000 and 1 x 107, between 100,000 and 5 x 106, or between 500,000 and 1 x 106. In some embodiments, the algorithms, models, regressors, and/or classifier of the present disclosure operate in a k-dimensional space, where k is a positive integer of 5 or greater (e.g., 5, 6, 7, 8, 9, 10, etc.). As such, the algorithms, models, regressors, and/or classifiers of the present disclosure cannot be mentally performed. [00169] As used herein, the term “untrained model” (e.g., “untrained classifier” and/or “untrained neural network”) refers to a machine learning model or algorithm, such as a classifier or a neural network, that has not been trained on a target dataset. In some embodiments, “training a model” (e.g., “training a neural network”) refers to the process of training an untrained or partially trained model (e.g., “an untrained or partially trained neural network”). Moreover, it will be appreciated that the term “untrained model” does not exclude the possibility that transfer learning techniques are used in such training of the untrained or partially trained model. For instance, Fernandes et al., 2017, “Transfer Learning with Partial Observability Applied to Cervical Cancer Screening,” Pattern Recognition and Image DB1/ 147040794.2 35 Attorney Ref. No.: 115834-5037-WO Analysis: 8th Iberian Conference Proceedings, 243-250, which is hereby incorporated by reference, provides non-limiting examples of such transfer learning. In instances where transfer learning is used, the untrained model described above is provided with additional data over and beyond that of the primary training dataset. Typically, this additional data is in the form of parameters (e.g., coefficients, weights, and/or hyperparameters) that were learned from another, auxiliary training dataset. Moreover, while a description of a single auxiliary training dataset has been disclosed, it will be appreciated that there is no limit on the number of auxiliary training datasets that can be used to complement the primary training dataset in training the untrained model in the present disclosure. For instance, in some embodiments, two or more auxiliary training datasets, three or more auxiliary training datasets, four or more auxiliary training datasets or five or more auxiliary training datasets are used to complement the primary training dataset through transfer learning, where each such auxiliary dataset is different than the primary training dataset. Any manner of transfer learning is used, in some such embodiments. For instance, consider the case where there is a first auxiliary training dataset and a second auxiliary training dataset in addition to the primary training dataset. In such a case, the parameters learned from the first auxiliary training dataset (by application of a first model to the first auxiliary training dataset) are applied to the second auxiliary training dataset using transfer learning techniques (e.g., a second model that is the same or different from the first model), which in turn results in a trained intermediate model whose parameters are then applied to the primary training dataset and this, in conjunction with the primary training dataset itself, is applied to the untrained model. Alternatively, in another example embodiment, a first set of parameters learned from the first auxiliary training dataset (by application of a first model to the first auxiliary training dataset) and a second set of parameters learned from the second auxiliary training dataset (by application of a second model that is the same or different from the first model to the second auxiliary training dataset) are each individually applied to a separate instance of the primary training dataset (e.g., by separate independent matrix multiplications) and both such applications of the parameters to separate instances of the primary training dataset in conjunction with the primary training dataset itself (or some reduced form of the primary training dataset such as principal components or regression coefficients learned from the primary training set) are then applied to the untrained model in order to train the untrained model. [00170] Furthermore, when a reference number is given an “ith” denotation, the reference number refers to a generic component, set, or embodiment. For instance, a model termed DB1/ 147040794.2 36 Attorney Ref. No.: 115834-5037-WO “model i” refers to the ith model in a plurality of models (e.g., a model 118-i in a plurality of models 118). In the present disclosure, unless expressly stated otherwise, descriptions of devices and systems will include implementations of one or more computers. [00171] In the present disclosure, unless expressly stated otherwise, descriptions of devices and systems will include implementations of one or more computers. For instance, and for purposes of illustration in Figure 1, 2A and 2B, a graphical data system 200 is represented as a single device that includes all the functionality of a computer system. Moreover, and for purposes of illustration in Figures 1 and 3, a client device 300 is represented as a single device that includes all the functionality of a computer system. However, the present disclosure is not limited thereto. For instance, in some embodiments, the functionality of the graphical data system 200 is spread across any number of networked computers and/or reside on each of several networked computers and/or by hosted on one or more virtual machines and/or containers at a remote location accessible across a communications network (e.g., communications network 186 of Figure 1). One of skill in the art will appreciate that a wide array of different computer topologies is possible for the graphical data system 200, and other devices and systems of the preset disclosure, and that all such topologies are within the scope of the present disclosure. Moreover, rather than relying on a physical communications network 186, the illustrated devices and systems may wirelessly transmit information between each other. As such, the exemplary topology shown in Figure 1 merely serves to describe the features of an embodiment of the present disclosure in a manner that will be readily understood to one of skill in the art. [00172] Figures 2A and 2B collectively depicts a block diagram of a graphical data system 200 according to some embodiments of the present disclosure. The graphical data system 200 at least facilitates optimizing encoding of graphical data, communication of graphical data, decoding of graphical data, or a combination thereof, such as by obtaining a training data set (e.g., block 404 of Figure 4A), using a first plurality of graphical data to train a feature extraction model (e.g., block 418 of Figure 4B), receiving a request to evaluate a second one or more graphical data (e.g., block 428 of Figure 4C), encoding (e.g., first sequence scan 1010-1 of Figure 11) an encoded byte stream for communication via a communication network 186 (e.g., block 442 of Figure 4D), matching a desired threshold performance to first training resolution (e.g., block 456 of Figure 4E), communicating the encoded byte stream (e.g., block 458 of Figure 4E), generating one or more instructions (e.g., DB1/ 147040794.2 37 Attorney Ref. No.: 115834-5037-WO block 458 of Figure 4E, block 466 of Figure 4E), terminating communication of the encoded byte stream (e.g., block 466 of Figure 4E), or a combination thereof. [00173] In some embodiments, the communication network 186 optionally includes the Internet, one or more local area networks (LANs), one or more wide area networks (WANs), other types of networks, or a combination of such networks. [00174] Examples of communication networks 186 include the World Wide Web (WWW), an intranet and/or a wireless network, such as a cellular telephone network, a wireless local area network (LAN) and/or a metropolitan area network (MAN), and other devices by wireless communication. The wireless communication optionally uses any of a plurality of communications standards, protocols and technologies, including Global System for Mobile Communications (GSM), Enhanced Data GSM Environment (EDGE), high-speed downlink packet access (HSDPA), high-speed uplink packet access (HSUPA), Evolution, Data-Only (EV-DO), HSPA, HSPA+, Dual-Cell HSPA (DC-HSPDA), long term evolution (LTE), near field communication (NFC), wideband code division multiple access (W-CDMA), code division multiple access (CDMA), time division multiple access (TDMA), Bluetooth, Wireless Fidelity (Wi-Fi) (e.g., IEEE 802.11a, IEEE 802.11ac, IEEE 802.11ax, IEEE 802.11b, IEEE 802.11g and/or IEEE 802.11n), voice over Internet Protocol (VoIP), Wi- MAX, a protocol for e-mail (e.g., Internet message access protocol (IMAP) and/or post office protocol (POP)), instant messaging (e.g., extensible messaging and presence protocol (XMPP), Session Initiation Protocol for Instant Messaging and Presence Leveraging Extensions (SIMPLE), Instant Messaging and Presence Service (IMPS)), and/or Short Message Service (SMS), or any other suitable communication protocol, including communication protocols not yet developed as of the filing date of this document. [00175] In various embodiments, the graphical data system 200 includes one or more processing units (CPUs) 172, a network or other communications interface 174, and memory 192. [00176] In some embodiments, the graphical data system 200 includes a user interface 176. The user interface 176 typically includes a display 178 for presenting media, such as a result by a plurality of models (e.g., first model 118-1, second model 118-2, ..., model X 118-X of Figure 2B) or a portion (e.g., some or all) of a plurality of graphical data (e.g., first graphical data 108-1 of Figure 1, second graphical data 108-2 of Figure 11, second graphical data 108- 2 of Figure 12, etc.). In some embodiments, the display 178 is integrated within the graphical DB1/ 147040794.2 38 Attorney Ref. No.: 115834-5037-WO data system 200 (e.g., housed in the same chassis as the CPU 172 and the memory 192). In some embodiments, the graphical data system 200 includes one or more input device(s) 180, which allow a subject to interact with the graphical data system 200. In some embodiments, the input devices 180 include a keyboard, a mouse, and/or other input mechanisms. Alternatively, or in addition, in some embodiments, the display 178 includes a touch- sensitive surface (e.g., where display 178 is a touch-sensitive display or the graphical data system 200 includes a touch pad). [00177] In some embodiments, the graphical data system 200 presents media to a user through the display 178. Examples of media presented by the display 178 include one or more images, a video, audio (e.g., waveforms of an audio sample), or a combination thereof. In typical embodiments, the one or more images, the video, the audio, or the combination thereof is presented by the display 178 through a client application 130. In some embodiments, the audio is presented through an external device (e.g., speakers, headphones, input/output (I/O) subsystem, etc.) that receives audio information from the graphical data system 200 and presents audio data based on this audio information. In some embodiments, the user interface 176 also includes an audio output device, such as speakers or an audio output for connecting with speakers, earphones, or headphones. [00178] The memory 192 includes high-speed random access memory, such as DRAM, SRAM, DDR RAM, or other random access solid state memory devices, and optionally also includes non-volatile memory, such as one or more magnetic disk storage devices, optical disk storage devices, flash memory devices, or other non-volatile solid state storage devices. The memory 192 may optionally include one or more storage devices remotely located from the CPU(s) 172. The memory 192, or alternatively the non-volatile memory device(s) within the memory 192, includes a non-transitory computer readable storage medium. Access to the memory 192 by other components of the graphical data system 200, such as the CPU(s) 172, is, optionally, controlled by a controller. In some embodiments, the memory 192 can include mass storage that is remotely located with respect to the CPU(s) 172. In other words, some data stored in memory 192 may in fact be hosted on devices that are external to the graphical data system 200, but that can be electronically accessed by the graphical data system 200 over an Internet, intranet, or other form of network 186 or electronic cable using communication interface 184. [00179] In some embodiments, the memory 192 of the graphical data system 200 optimizing decoding of graphical data 108 stores: DB1/ 147040794.2 39 Attorney Ref. No.: 115834-5037-WO • optionally, an operating system 102 (e.g., ANDROID, iOS, DARWIN, RTXC, LINUX, UNIX, OS X, WINDOWS, or an embedded operating system such as VxWorks) that includes procedures for handling various basic system services; • optionally, an electronic address 104 associated with the graphical data system 200 that identifies graphical data system 200 (e.g., within the communication network 186); • a graphical data store 106 that stores a record of graphical data (e.g., first plurality of graphical data 108-1, second one or more graphical data 108-2, ..., graphical data C 108- C of Figure 2A), each graphical data 108 is defined by a plurality of characteristics (e.g., first plurality of characteristics 110-1 of Figure 2A), that collectively characterize a corresponding graphical data 108 by one or more characteristics (e.g., value of first characteristic 112-1-1, value of second characteristic 112-2-1, ..., value of characteristic P 112-1-P of first plurality of characteristics 110-1 of Figure 2A) that is utilized by one or more models 118, such as for obtaining a training data set, obtaining a trained feature extraction model 118, and, optionally, an encoded byte stream including a plurality of sequence scans associated with the corresponding graphical data 108; • a model library 116 that retains a plurality of models (e.g., first model 118-1, second model 118-2, ..., model F 118-F of Figure 2B), in which each respective model 118 is configured for evaluating, at least in part, a plurality of graphical data 108 in accordance with one or more parameters of the model 118 (e.g., first parameter 120-1, second parameter 120- 2, ..., parameter D 120-D of first model 118-1 of Figure 2B); • a codec module 122 that retains one or more codecs (e.g., first codec 124-1, second codec 124-2, ..., codec H 124-H of Figure 2B), each codec 124 in one or more of codecs 124 including one or more instructions that is specific for converting a respective graphical data 108 into an encoded byte stream (e.g., second encoded byte stream 114-2 of second graphical data 108-2 of Figure 2A) in accordance with a corresponding encoder (e.g., first encoder 126- 1, second encoder 126-2, ..., encoder H 126-H of Figure 2B, etc.), reconstructing the respective graphical data 108 from the encoded byte stream 114 in accordance with a corresponding decoder (e.g., first decoder 128-1, second decoder 126-2, ..., decoder H 128- H of Figure 2B, etc.), generating one or more instructions (e.g., a first instruction to terminate decoding of the first encoded byte stream 114-1 at a first client device 300-1, a second instruction to terminate communication of the encoded bye stream 114-2 at the graphical data system 300, etc.), or a combination thereof; and DB1/ 147040794.2 40 Attorney Ref. No.: 115834-5037-WO • optionally, a client application 130 for presenting information (e.g., media) using a display 178 of the graphical data system 200. [00180] As indicated above, preferably, the graphical data system 200 includes an operating system 102 that includes procedures for handling various basic system services. The operating system 102 (e.g., iOS, ANDROID, DARWIN, RTXC, LINUX, UNIX, OS X, WINDOWS, or an embedded operating system such as VxWorks) includes various software components and or drivers for controlling and managing general system tasks (e.g., memory management, storage device control, power management, etc.) and facilitates communication between various hardware and software components. [00181] In some embodiments, an optional electronic address 104 is associated with the graphical data system 200. The optional electronic address 104 is utilized to at least uniquely identify the graphical data system 200 from other devices and components of the distributed system 100, such as other client devices 300 having access to the communications network 186. For instance, in some embodiments, the electronic address 104 is utilized to receive a request from a client device 300 to communicate a graphical data 108 set through an encoded byte stream 114 in the form of a first codec 124-1. However, the present disclosure is not limited thereto. [00182] The graphical data system 200 includes a graphical data store 106 that stores a variety of graphical data 108, such as one or more graphical data sets (e.g., first plurality of graphical data 108-1, second one or more graphical data 108-2, ..., plurality of graphical data C 108-C of Figure 2A). For instance, in some embodiments, the graphical data store 106 is configured to retain between three and 20 graphical data 108 sets, between 5 and 40 graphical data sets, between 15 and 100 graphical data sets, or between 25 and 150 graphical data sets. In some embodiments, the graphical data store 106 is configured to retain between three and 20 graphical data sets 108, between 5 and 40 graphical data sets, between 15 and 100 graphical data sets, or between 25 and 150 graphical data sets. In some embodiments, the graphical data store 106 is configured to retain at least 5 graphical data sets 108, at least 15 graphical data sets, at least 25 graphical data sets, at least 50 graphical data sets, at least 100 graphical data sets, or at least 150 graphical data sets. In some embodiments, the graphical data store 106 is configured to retain at most 5 graphical data sets 108, at most 15 graphical data sets, at most 25 graphical data sets, at most 50 graphical data sets, at most 100 graphical data sets, or at most 150 graphical data sets. DB1/ 147040794.2 41 Attorney Ref. No.: 115834-5037-WO [00183] Each respective graphical data 108 is defined by a corresponding plurality of characteristics (e.g., first plurality of characteristics 110-1 define first plurality of graphical data 108-1 of Figure 2A). For instance, in some embodiments, each respective plurality of graphical data 108 is defined by between three and 20 characteristics 112, between 5 and 40 characteristics, between 15 and 100 characteristics, or between 25 and 150 characteristics. In some embodiments, each respective plurality of graphical data 108 is defined by between three and 20 characteristics 112, between 5 and 40 characteristics, between 15 and 100 characteristics, or between 25 and 150 characteristics. In some embodiments, each respective plurality of graphical data 108 is defined by at least 5 characteristics 112, at least 15 characteristics, at least 25 characteristics, at least 50 characteristics, at least 100 characteristics, or at least 150 characteristics. In some embodiments, each respective plurality of graphical data 108 is defined by at most 5 characteristics 112, at most 15 characteristics, at most 25 characteristics, at most 50 characteristics, at most 100 characteristics, or at most 150 characteristics 112. [00184] Collectively, the plurality of characteristics 110 characterize the corresponding graphical data 108 by one or more characteristics 112. The one or more characteristics provide information about one or more aspects of the corresponding graphical data 108. For instance, in some embodiments, the one or more characteristics 112 includes a first characteristic 112-1 associated with a corresponding modality of the plurality of graphical data 108 (e.g., a first value of the first characteristic 112-1 is associated with a first modality, a second value of the first characteristic 112-1 is associated with a second modality different than the first modality, etc.), a second characteristic 112-2 associated with a sampling resolution of the plurality of graphical data 108, a third characteristic 112-3 associated with a file format of the plurality of graphical data, a fourth characteristic 112-4 associated with a subject matter (e.g., content) of the plurality of graphical data, a fifth characteristic 112-5 associated with a capture setting (e.g., camera setting) associated with the plurality of graphical data, or a combination thereof. However, the present disclosure is not limited thereto. For instance, in some embodiments, a respective characteristic 112 in the plurality of characteristics 112 associated with the plurality of graphical data 108 is utilized as input for a model 118, such as for identifying a first resolution in a plurality of resolutions for optimizing decoding of the plurality of graphical data 108 and/or determining if a corresponding desired threshold performance is satisfied. However, the present disclosure is not limited thereto. DB1/ 147040794.2 42 Attorney Ref. No.: 115834-5037-WO [00185] For instance, in some embodiments, the third characteristic 112-3 associated with the file format of the plurality of graphical data 108 allows for retaining graphical data in any electronic image file format, including but not limited to JPEG/JFIF, TIFF, Exif, PDF, EPS, GIF, BMP, PNG, PPM, PGM, PBM, PNM, WebP, HDR raster formats, HEIF, BAT, BPG, DEEP, DRW, ECW, FITS, FLIF, ICO, ILBM, IMG, PAM, PCX, PGF, JPEG XR, Layered Image File Format, PLBM, SGI, SID, CD5, CPT, PSD, PSP, XCF, PDN, CGM, SVG, PostScript, PCT, WMF, EMF, SWF, XAML, and/or RAW. Furthermore, in some such embodiments, the ability of the graphical data system 200 to retain a variety of electronic allows for a respective codec 124 to compress and reconstruct the plurality of graphical data 108 in accordance with a corresponding encoder 126 and a corresponding decoder 128 with more than one file type, such as with two-dimensional (2D) graphical data, three-dimensional (3D) graphical data, four-dimensional (4D) graphical data, or a combination thereof. [00186] In some embodiments, the plurality of graphical data 108 is obtained in any electronic color mode, including but not limited to grayscale, bitmap, indexed, RGB, CMYK, HSV, lab color, duotone, and/or multichannel. In some embodiments, the plurality of graphical data 108 is manipulated (e.g., stitched, compressed and/or flattened). In some embodiments, the plurality of graphical data 108 has a file size is between 1 KB and 1 MB, between 1 MB and 0.5 GB, between 0.5 GB and 5 GB, between 5 GB and 10 GB, between 10 GB and 500 GB, or greater. In some embodiments, the plurality of graphical data 108 has a file size of at least 1 KB, a file size of at least 1 MB, a file size of at least 0.5 GB, or a file size of at least 10 GB. [00187] In some embodiments, the plurality of graphical data 108 includes between 1 million and 25 million pixels. In some embodiments, each resolution is represented by five or more, ten or more, 100 or more, 1,000 or more contiguous pixels in an image. In some embodiments, each resolution is represented by between 1,000 and 250,000 contiguous pixels in a native image 125. [00188] In some embodiments, the plurality of graphical data 108 is represented as an array (e.g., matrix) including a plurality of pixels, such that the location of each respective pixel in the plurality of pixels in the array (e.g., matrix) corresponds to its original location in the image. In some embodiments, the plurality of graphical data 108 is represented as a vector including a plurality of pixels, such that each respective pixel in the plurality of pixels in the vector includes spatial information corresponding to its original location in the image. DB1/ 147040794.2 43 Attorney Ref. No.: 115834-5037-WO [00189] In some embodiments, a pixel includes one or more pixel values (e.g., an intensity value). In some embodiments, each respective pixel in the plurality of pixels includes one pixel intensity value, such that the plurality of pixels represents a single-channel image including a one-dimensional integer vector including the respective pixel values for each respective pixel. For example, in some embodiments, an 8-bit single-channel image (e.g., grey-scale) includes 28 or 256 different pixel values (e.g., 0-255). In some embodiments, each respective pixel in the plurality of pixels of an image includes a plurality of pixel values, such that the plurality of pixels represents a multi-channel image including a multi- dimensional integer vector, where each vector element represents a plurality of pixel values for each respective pixel. For example, in some embodiments, a 24-bit 3-channel image (e.g., RGB color) includes 224 (e.g., 28x3) different pixel values, where each vector element includes 3 components, each between 0-255. In some embodiments, an n-bit image of the plurality of graphical data includes up to 2n different pixel values, where n is any positive integer. See, Uchida, 2013, “Image processing and recognition for biological images,” Develop. Growth Differ., 55, pg.523-549, doi:10.1111/dgd.12054, which is hereby incorporated herein by reference in its entirety for all purposes. [00190] Referring to Figure 2B, the graphical data system 200 includes a model library 116 that stores a plurality of models 118 (e.g., classifiers, regressors, clustering, etc.). In some embodiments, the model library 116 stores two more models 118 (e.g., a first feature extraction model 118-1 and a second feature extraction model 118-2), three or more models (e.g., a first segmentation model 118-1, a second first classification model 118-2, a third second classification model 118-3), four or more models, ten or more models, 50 or more models, or 100 or more models. In some embodiments, a model 118 in the plurality of models 118 of the model library 116 is implemented as an artificial intelligence engine. For instance, in some embodiments, the model 118 includes one or more feature extraction models including one or more segmentation models 118, classification models 118, one or more gradient boosting models 118, one or more random forest models 118, one or more neural network (NN) models 118, one or more regression models, one or more Naïve Bayes models 118, one or more machine learning algorithms (MLA) 118, or a combination thereof. In some embodiments, an MLA or a NN is trained from a training data set that includes one or more features identified from a data set. MLAs include supervised algorithms (such as algorithms where the features/classifications in the data set are annotated) using linear regression, logistic regression, decision trees, classification and regression trees, Naïve DB1/ 147040794.2 44 Attorney Ref. No.: 115834-5037-WO Bayes, nearest neighbor clustering; unsupervised algorithms (such as algorithms where no features/classification in the data set are annotated a priori), such as means clustering, principal component analysis, random forest, adaptive boosting; and semi-supervised algorithms (such as algorithms where an incomplete number of features/classifications in the data set are annotated) using generative approach (such as a mixture of Gaussian distributions, mixture of multinomial distributions, hidden Markov models), low density separation, graph-based approaches (such as minimum cut, harmonic function, manifold regularization, etc.), heuristic approaches, or support vector machines. [00191] Neural network models 118 include conditional random fields models 118, convolutional neural network (CNN) models 118, attention based neural network models 118, deep learning models 118, long short term memory network model 118, or other neural models 118. [00192] While MLA and neural networks identify distinct approaches to machine learning, the terms may be used interchangeably herein. Thus, a reference to MLA may include a corresponding NN or a reference to NN may include a corresponding MLA unless explicitly stated otherwise. In some embodiments, the training of a respective model 118 includes providing one or more optimized data sets, such as a first plurality of graphical data 108-1, labeling features as they occur (e.g., as a characteristic 112 associated with the plurality of graphical data 108), and training the MLA to predict or classify based on new inputs. Artificial NNs are efficient computing models 118 which have shown their strengths in solving hard problems in artificial intelligence. For instance, artificial NNs have also been shown to be universal approximators, that is, they can represent a wide variety of functions when given appropriate parameters. [00193] One of skill in the art will readily appreciate other models 118 that are applicable to the systems and methods of the present disclosure. In some embodiments, the systems and methods of the present disclosure utilize more than one model 118 to provide an evaluation (e.g., arrive at an evaluation given one or more inputs), such as an identity of a first resolution with an increased accuracy. For instance, in some embodiments, each respective model 118 arrives at a corresponding evaluation when provided a respective data set. Accordingly, in some embodiments, each respective model 118 independently arrives at a result and then the result of each respective model 118 is collectively verified through a comparison or amalgamation of the models 118. From this, a cumulative result is provided by the models 118. However, the present disclosure is not limited thereto. DB1/ 147040794.2 45 Attorney Ref. No.: 115834-5037-WO [00194] In some embodiments, a respective model 118 is tasked with performing a corresponding activity. As a non-limiting example, in some embodiments, the task performed by the respective model 118 includes, but is not limited to, evaluating a training data set (e.g., first plurality of graphical data 108-1 of Figure 2A, block 402 of Figure 4A), determining a desired threshold performance (e.g., block 428 of Figure 4C), determining if the desired threshold performance is satisfied (e.g., block 456 of Figure 4E, block 458 of Figure 4E, block 466 of Figure 4E), matching the desired threshold performance to a training resolution (e.g., block 456 of Figure 4E)), generating one or more instructions for terminating decoding of the encoded byte stream at a client device (e.g., block 458 of Figure 4E, block 466 of Figure 4E, etc.), any combination thereof. [00195] In some embodiments, each respective model 118 of the present disclosure makes use of 10 or more parameters, 100 or more parameters, 1000 or more parameters, 10,000 or more parameters, or 100,000 or more parameters. In some embodiments, each respective model 118 of the present disclosure cannot be mentally performed. [00196] In some embodiments, the codec module 122 is configured to retain a plurality of codecs (e.g., first codec 124-1, second codec 124-2, ..., codec H 124-H of Figure 2B). For instance, in some embodiments, the codec module 122 is configured to retain a plurality of codecs including between three and 20 codecs 124, between 5 and 40 codecs, between 15 and 100 codecs, or between 25 and 150 codecs. In some embodiments, the codec module 122 is configured to retain between three and 20 codecs 124, between 5 and 40 codecs, between 15 and 100 codecs, or between 25 and 150 codecs. In some embodiments, the codec module 122 is configured to retain at least 5 codecs, at least 15 codecs, at least 25 codecs, at least 50 codecs, at least 100 codecs, or at least 150 codecs. In some embodiments, the codec module 122 is configured to retain at most 5 codecs, at most 15 codecs, at most 25 codecs, at most 50 codecs, at most 100 codecs, or at most 150 codecs. [00197] In some embodiments, each codec 124 includes one or more specific instructions for converting graphical data 108 into an encoded byte stream (e.g., encoded byte stream 114 of Figure 2A) in accordance with a corresponding encoder (e.g., first encoder 126-1, second encoder 126-2, ..., encoder H 126-H of Figure 2B, etc.) and reconstructing the plurality of graphical data from the encoded byte stream in accordance with a corresponding decoder (e.g., first decoder 128-1, second decoder 126-2, ..., decoder H 128-H of Figure 2B, etc.). DB1/ 147040794.2 46 Attorney Ref. No.: 115834-5037-WO [00198] In some embodiments, after encoding a respective graphical data 108 into an encoded byte stream 114, the codec module 122 retains the encoded coded byte stream 114, which removes a requirement to encode the respective graphical data 108 each time a request for the respective graphical data 108 is received at the graphical data system 200. However, the present disclosure is not limited thereto. [00199] For instance, in some embodiments, a respective encoder 126 includes one or more instructions to encode (e.g., progressively encode) a respective plurality of graphical data 108 into an encoded byte stream 114 through a sampling resolution of the respective plurality of graphical data. The encoded byte stream 114 includes a plurality of sequence scans, which is independent of a corresponding modality and/or a corresponding resolution (e.g., sampling resolution) property 112 associated with the respective plurality of graphical data 108. As a non-limiting example, in some embodiments, a first plurality of graphical data 108-1 includes M 2D images with native resolutions Xi ∀i ∈ {1,2, ... ,M}. Accordingly, in some embodiments, the respective encoder 126 imposes a condition that the native resolution of an initial terminal scan in the plurality of sequence scans exceeds a minimum of X (Xmin), in order to recover sufficient information from evaluation of the initial terminal scan. [00200] In some embodiments, a respective decoder 128 includes one or more instructions to decode an encoded byte stream 114 into a respective scan in a plurality of sequence scans of the encoded byte stream 114 with a corresponding native resolution. As a non-limiting example, if only 2% of the encoded byte stream is communicated to the respective decoder via a communication network 186, the respective decoder 128 is able to decode the encoded byte stream and complete reconstruction of a respective plurality of graphical data associated with the encoded byte stream. For instance, referring briefly to Figures 5 and 11, in some embodiments, a high-throughput JPEG 2000 (“HTJ2K”) codec (e.g., first codec 124-1 of Figure 2B) is an extension to a JPEG 2000 (“J2K”) codec (e.g., second codec 124-2 of Figure 2B), that, unlike an original JPEG standard codec (e.g., third codec 124-3 of Figure 2B), enables faster decoding (e.g., using first decoder 128-1), higher quality, lossless encoding with high scalability (e.g., using first encoder 126-1), and progressive decoding of images through an encoding (e.g., compression) and decoding (e.g., decompression) of a plurality of graphical data 108. See Foos, et al., 2000, “JPEG 2000 Compression of Medical Imagery,” Medical Imaging 2000: PACS Design and Evaluation: Engineering and Clinical Issues, 3980, pg.85-96, SPIE; Taubman, et al., 2019, “High Throughput JPEG 2000 (HTJ2K) Algorithm, DB1/ 147040794.2 47 Attorney Ref. No.: 115834-5037-WO Performance and Potential,” International Telecommunications Union (ITU), pg.15444-15, each of which is hereby incorporated by reference in its entirety for all purposes. [00201] The HTJ2K codec 124 utilizes a lightweight and efficient block-coder model that yields up to a factor of 10-30 increase in throughput for lossless coding and a 6-10% increase in efficiency when compared to J2K. See Taubman et al., 2020, “High Throughput JPEG 2000 (HTJ2K) and the JPH File Format: A Primer,” available at ds.jpeg.org/whitepapers/jpeg-htj2k-whitepaper.pdf (accessed April 17, 2023), which is hereby incorporated by reference its entirety for all purposes. Furthermore, HTJ2K’s lower computational complexity and support for parallelism results in a throughput on par, or even better, than the heavily optimized original JPEG standard in single and multi-threaded applications. See Taubman, et al., 2019. For instance, in some embodiments, the HTJ2K codec 124 is utilized in a workflow by extending a DICOM transfer syntax standard to add support for HTJ2K. See Hafey et al., 2022, “HTJ2K Transfer Sytax Support,” available at dicomstandard.org/docs/librariesprovider2/default-documentlibrary/htj2k-transfer-syntax- support.docx (accessed April 17, 2023). [00202] In some embodiments, a client application 130 is a group of instructions that, when executed by the processor 174, generates content for presentation to the user (e.g., graphical data 108-1 of Figure 11, etc.), such as a result provided by one or more models 118. In some embodiments, the client application 130 generates content in response to one or more inputs received from the user through the graphical data system 200, such as the inputs 180 of the graphical data system 200. [00203] Each of the above identified modules and applications correspond to a set of executable instructions for performing one or more functions described above and the methods described in the present disclosure (e.g., the computer-implemented methods and other information processing methods described herein). These modules (e.g., sets of instructions) need not be implemented as separate software programs, procedures or modules, and thus various subsets of these modules are, optionally, combined or otherwise re-arranged in various embodiments of the present disclosure. In some embodiments, the memory 192 optionally stores a subset of the modules and data structures identified above. Furthermore, in some embodiments, the memory 192 stores additional modules and data structures not described above. DB1/ 147040794.2 48 Attorney Ref. No.: 115834-5037-WO [00204] It should be appreciated that the graphical data system of Figures 2A and 2B is only one example of a graphical data system, and that the graphical data system 200 optionally has more or fewer components than shown, optionally combines two or more components, or optionally has a different configuration or arrangement of the components. The various components shown in Figures 2A and 2B are implemented in hardware, software, firmware, or a combination thereof, including one or more signal processing and/or application specific integrated circuits. [00205] Referring to Figure 3, a description of an exemplary client device 300 that can be used with the present disclosure is provided. In some embodiments, a client device 300 includes a smart phone (e.g., an iPhone, an Android device, etc.), a laptop computer, a tablet computer, a desktop computer, a wearable device (e.g., a smart watch, a heads-up display (HUD) device, etc.), a television (e.g., a smart television), or another form of electronic device such as a gaming console, a stand-alone device, and the like. [00206] The client device 300 illustrated in Figure 3 has one or more processing units (CPU’s) 272, a network or other communications interface 274, a memory 292 (e.g., random access memory), a user interface 276, the user interface 276 including a display 278 and input 280 (e.g., keyboard, keypad, touch screen, etc.), optional audio circuitry, an optional speaker, an optional microphone, an optional input/output (I/O) subsystem, one or more communication busses 270 for interconnecting the aforementioned components, and a power system (e.g., power supply) for powering the aforementioned components. [00207] In some embodiments, the input 280 is a touch-sensitive display, such as a touch- sensitive surface. In some embodiments, the user interface 276 includes one or more soft keyboard embodiments. In some embodiments, the soft keyboard embodiments include standard (QWERTY) and or non-standard configurations of symbols on the displayed icons. The input 280 and/or the user interface 276 is utilized by an end-user of the respective client device 300 (e.g., a respective subject) to input various information (e.g., a text object within a message) to the respective client device. [00208] In some embodiments, the client device 300 illustrated in Figure 3 optionally includes, in addition to accelerometer(s), a magnetometer, and a global positioning system (GPS or GLONASS or other global navigation system) receiver for obtaining information concerning a current location (e.g., a latitude, a longitude, an elevation, etc.) and/or an orientation (e.g., a portrait or a landscape orientation of the device) of the client device 300. DB1/ 147040794.2 49 Attorney Ref. No.: 115834-5037-WO In some embodiments, the location of the client device sets a respective codec 124 for encoding a respective graphical data 108. However, the present disclosure is not limited thereto. [00209] It should be appreciated that the client device 300 illustrated in Figure 3 is only one example of a multifunctional device that may be used for receiving an encoded byte stream 114, decoding the encoded byte stream 114, communicating a request for an evaluation of graphical data 108, terminating decoding of the encoded byte stream 114, or a combination thereof. Thus, the client device 300 optionally has more or fewer components than shown, optionally combines two or more components, or optionally has a different configuration or arrangement of the components. The various components shown in Figure 3 are implemented in hardware, software, firmware, or a combination thereof, including one or more signal processing and/or application specific integrated circuits. In some embodiments client device 300 is a desktop or laptop computer. [00210] The memory 292 of the client device 300 illustrated in Figure 3 optionally includes high-speed random access memory and optionally also includes non-volatile memory, such as one or more magnetic disk storage devices, flash memory devices, or other non-volatile solid- state memory devices. Access to the memory 292 by other components of the client device 300, such as CPU(s) 272 is, optionally, controlled by the memory controller. [00211] In some embodiments, the one or more CPU(s) 272 run or execute various software programs and/or sets of instructions stored in the memory 292, such as the client application 230, to perform various functions for the client device 300 and process data. [00212] In some embodiments, the CPU(s) 272 and the memory controller are implemented on a single chip. In some other embodiments, the CPU(s) 272 and the memory controller are implemented on separate chips. [00213] In some embodiments, the audio circuitry, the optional speaker, and the optional microphone provide an audio interface between the respective subject and the client device 300, enabling the client device to provide a message that include audio data provided through the audio circuitry, the optional speaker, and/or the optional microphone. The audio circuitry receives audio data from the peripherals interface, converts the audio data to electrical signals, and transmits the electrical signals to the speaker. The speaker converts the electrical signals to human-audible sound waves. The audio circuitry also receives electrical signals converted by the microphone from sound waves. The audio circuitry converts the electrical DB1/ 147040794.2 50 Attorney Ref. No.: 115834-5037-WO signal to audio data and transmits the audio data to peripherals interface for processing. Audio data is, optionally, retrieved from and or transmitted to the memory 292 and or the RF circuitry by the peripherals interface. [00214] In some embodiments, the client device 300 optionally also includes one or more optical sensors. The optical sensor(s) optionally include charge-coupled device (CCD) or complementary metal-oxide semiconductor (CMOS) phototransistors. The optical sensor(s) receive light from the environment, projected through one or more lens, and converts the light to data representing an image. The optical sensor(s) optionally capture still images and or video. In some embodiments, an optical sensor is disposed on a back end portion of the client device 300 (e.g., opposite the display 278 on a front end portion of the client device 300) so that the input 280 is enabled for use as a viewfinder for still (e.g., digital image, etc.) and or video graphical data acquisition. [00215] In some embodiments, the memory 292 of the client device 300 stores: • an operating system 202 that includes procedures for handling various basic system services; • an electronic address 204 associated with the client device 300; • a model library 216 that retains a plurality of models (e.g., first model 118-1, second model 118-2, ..., model F 118-X of Figure 2B), in which each respective model 118 is configured for evaluating, at least in part, a plurality of graphical data 108 in accordance with one or more parameters of the model 118 (e.g., first parameter 120-1, second parameter 120- 2, ..., parameter D 120-D of first model 118-1 of Figure 2B); • a codec module 222 that retains a plurality of codecs (e.g., first codec 124-1, second codec 124-2, ..., codec H 124-H of Figure 2B), each codec 124 in the plurality of codecs 124 including specific instructions for converting graphical data 108 into an encoded byte stream in accordance with a corresponding encoder (e.g., first encoder 126-1, second encoder 126- 2, ..., encoder H 126-H of Figure 2B, etc.), reconstructing the plurality of graphical data from the encoded byte stream in accordance with a corresponding decoder (e.g., first decoder 128-1, second decoder 126-2, ..., decoder H 128-H of Figure 2B, etc.), generating one or more instructions (e.g., a first instruction to terminate decoding of the first encoded byte stream 114-1 at a first client device 300-1, a second instruction to terminate communication of the encoded bye stream 114-2 at the graphical data system 300, etc.), or a combination thereof; and DB1/ 147040794.2 51 Attorney Ref. No.: 115834-5037-WO • optionally, a client application 130 for presenting information (e.g., media) using a display 278. [00216] As illustrated in Figure 3, a client device 300 preferably includes an operating system 202 that includes procedures for handling various basic system services. The operating system 202 (e.g., iOS, ANDROID, DARWIN, RTXC, LINUX, UNIX, OS X, WINDOWS, or an embedded operating system such as VxWorks) includes various software components and or drivers for controlling and managing general system tasks (e.g., memory management, storage device control, power management, etc.) and facilitates communication between various hardware and software components. [00217] An electronic address 204 is associated with each client device 300, which is utilized to at least uniquely identify the client device from other devices and components of the system 100. In some embodiments, the electronic address 204 of the client device 300 has the same functionality as the electronic address 104 of the graphical data system 200. However, the present disclosure is not limited thereto. [00218] In some embodiments, the client device includes a model library 216 that stores a plurality of models 118 (e.g., classifiers, regressors, clustering, etc.). In some embodiments, the model library 216 of the client device 300 has the same functionality as the model library 116 of the graphical data system 200. However, the present disclosure is not limited thereto. [00219] In some embodiments, the client device includes a codec module 222 that stores a plurality of codecs 124. In some embodiments, the codec module 222 of the client device 300 has the same functionality as the codec module 122 of the graphical data system 200. However, the present disclosure is not limited thereto. [00220] In some embodiments, the client application 230 is a group of instructions that, when executed by the processor 272, generates content for presentation to the respective subject (e.g., graphical data set 108-2 of Figure 11), such as a result of a corresponding report generated by one or more models 118. In some embodiments, the client application 230 generates content in response to one or more inputs received from the respective subject through the client device 300, such as the inputs 280 of the client device 300. In some embodiments, the client application 230 of the client device 300 has the same functionality as the client application 130 of the graphical data system 200. However, the present disclosure is not limited thereto. DB1/ 147040794.2 52 Attorney Ref. No.: 115834-5037-WO [00221] In some embodiments, the client device 300 has any or all of the circuitry, hardware components, and software components found in the system depicted in Figure 3. In the interest of brevity and clarity, only a few of the possible components of the client device 300 are shown to better emphasize the additional software modules that are installed on the client device. [00222] Each of the above identified modules and applications correspond to a set of executable instructions for performing one or more functions described above and the methods described in the present disclosure. These modules (e.g., sets of instructions) need not be implemented as separate software programs, procedures or modules, and thus various subsets of these modules are, optionally, combined or otherwise re-arranged in various embodiments of the present disclosure. In some embodiments, the memory 292 optionally stores a subset of the modules and data structures identified above. Furthermore, in some embodiments, the memory 292 stores additional modules and data structures not described above. [00223] It should be appreciated that Figure 3 illustrates only an example of the client device 300, and that the client device 300 optionally has more or fewer components than shown, optionally combines two or more components, or optionally has a different configuration or arrangement of the components. The various components shown in Figure 3 are implemented in hardware, software, firmware, or a combination thereof, including one or more signal processing and/or application specific integrated circuits. Moreover, the client device 300 can be a single device that includes all the functionality of the client device 300. The client device 300 can also be a combination of multiple devices. For instance, the functionality of the client device 300 may be spread across any number of networked computers and/or reside on each of several networked computers and/or by hosted on one or more virtual machines and/or containers at a remote location accessible across a communications network (e.g., communications network 186, network interface 284, or both). One of skill in the art will appreciate that a wide array of different computer topologies is possible for the client device 300, and other devices and systems of the preset disclosure, and that all such topologies are within the scope of the present disclosure. [00224] Now that a general topology of the distributed system 100 has been described in accordance with various embodiments of the present disclosures, details regarding some processes in accordance with Figures 4A through 4E will be described. DB1/ 147040794.2 53 Attorney Ref. No.: 115834-5037-WO [00225] Various modules in a memory 192 of a graphical data system 200 (e.g., graphical data store 106 of Figure 2A, model library 116 of Figure 2B, codec module 122 of Figure 2B, etc.) and/or a memory 292 of a client device 300200 (e.g., model library 216 of Figure 3, codec module 222 of Figure 3, etc.) perform certain processes of the methods of the present disclosure, unless expressly stated otherwise. Furthermore, it will be appreciated that the processes of a method of the present disclosure can be encoded in a single module or any combination of modules. [00226] Referring now to Figures 4A through 4E, there is depicted a flowchart illustrating an exemplary method 400 in accordance with some embodiments of the present disclosure. In the flowchart, the preferred parts of the method are shown in solid line boxes, whereas additional, optional, or alterative parts of the method are shown in dashed line boxes. [00227] In some embodiments, the method 400 optimizes encoding of graphical data (e.g., graphical data 108 of Figure 2A), communication of graphical data (e.g., encoded byte stream 114 of Figure 2A), decoding of graphical data, or a combination thereof that is communicated via a communication network (e.g., communication network 186 of Figure 1). In particular, in some embodiments, the method 400 communicates an encoded byte stream (e.g., second encoded byte stream 114-2 of Figure 2A) in accordance with a prior training using a different graphical data set (e.g., first graphical data 108-1 of Figure 2A) associated with the encoded byte stream, a use case defined by a request for the encoded byte stream 114 (e.g., defined by a request received from a second computer system such as client device 300 of Figure 1), an aspect of a client device 300 associated with the request for the encoded byte stream 114, a required resolution for using the graphical data 108 of the encoded byte stream 114, an aspect of a model (e.g., model 118 of Figure 2B) to which the graphical data 108 is applied, or a combination thereof. From this, in some embodiments, the method 400 allows for communicating only an optimal resolution, or corresponding scan in a plurality of sequence scans of the encoded byte stream 114 that matches a desired threshold performance. Accordingly, in some embodiments, the method 400 allows or optimizing cost (e.g., data storage costs, data capacity costs, etc.), network bandwidth, turnaround time, or a combination thereof for communicating graphical information through the encoded byte stream 114. Moreover, in some embodiments, the method 400 allows for terminating communication of the encoded byte stream in accordance with a determination that a first resolution of the encoded byte stream decoded at the second computer system satisfies a DB1/ 147040794.2 54 Attorney Ref. No.: 115834-5037-WO threshold resolution, such as by matching or exceeding the first training resolution. However, the present disclosure is not limited thereto. [00228] Block 402. Referring to block 402 of Figure 4A, in various embodiments, the method 400 is provided at a computer system (e.g., graphical data system 200 of Figures 2A and 2B, client device 300 of Figure 3, etc.). The computer system 200 includes one or more processors (e.g., CPU 172 of Figure 2A, CPU 272 of Figure 3, etc.) and a memory (e.g., memory 192 of Figures 2A and 2B, memory 392 of Figure 3, etc.), such a first memory coupled to the one or more processors. In some embodiments, the memory includes one or more programs configured to be executed by the one or more processors (e.g., graphical data store 106 of Figure 2A, model library 116 of Figure 2B, codec module 122 of Figure 2B, model library 216 of Figure 3, codec module 222 of Figure 3, etc.). Accordingly, in some such embodiments, the method 400 requires utilization of a computer system, such as in order to evaluate the graphical data 108 and communicate the graphical data 108 via the communication network 186, and, therefore, cannot be mentally performed. [00229] Furthermore, in some embodiments, the method 400 allows for parallelism, which allows for encoding (e.g., using first encoder 126-1 of Figure 2B), communicating an encoded byte stream, decoding the encoded byte stream (e.g., using second decoder 126-2 of Figure 2B), or a combination thereof simultaneously using a plurality of graphical data 108 using a plurality of CPUs 172. However, the present disclosure is not limited thereto. [00230] Block 404. Referring to block 404, in some embodiments, the method 400 includes obtaining, in electronic form, a training data set. In some embodiments, the obtaining is performed by one or more models (e.g., first model 118-2, second model 118-2, model 118 F 118-F of Figure 2B, etc.), which utilize the training data as input features. For instance, in some embodiments, the one or more models 118 include at least one feature extraction model 118, such as a first segmentation model 118-1 or a second classification model 118-2. However, the present disclosure is not limited thereto. In some embodiments, the obtaining the training data set includes accessing a respective plurality of graphical data 108 associated with the first computer system. For instance, in some embodiments, the training data set is obtained by accessing the first graphical data 108-1 stored locally at the first computer system. However, the present disclosure is not limited thereto. In some embodiments, the training data set is obtained by accessing the first graphical data 108-1 that is stored remotely from the first computer system (e.g., stored at a first client device 300-1 of Figure 1). DB1/ 147040794.2 55 Attorney Ref. No.: 115834-5037-WO [00231] In some embodiments, the training data set is selected, at least in part, in accordance with an aspect of a respective model 118 the training data set is applied to, such as selecting a first training data set for application with a first model 118-1 and a second training data set for application with a second model 118-2. In some embodiments, the training data is selected, at least in part, in accordance with an aspect of a data type of the training data, such as such as by using the first training data set for application when an evaluation is associated with a first modality (e.g., an X-Ray modality) and the second training data set is used for application when the evaluation is associated with a second modality (e.g., an MRI modality). However, the present disclosure is not limited thereto. [00232] In some embodiments, the training data set includes a first plurality of graphical data (e.g., first plurality of graphical data 108-1 of Figure 2A). The first plurality of graphical data 108-1 is defined by a first plurality of characteristics (e.g., first characteristics 110-1 of Figure 2A) that includes one or more characteristics (e.g., first characteristic 112-1 of Figure 2A, third characteristic 112-3 of Figure 2A). In some embodiments, the one or more characteristics 112 includes a sampling resolution (e.g., first characteristic 112-1 of Figure 2A) of the first plurality of graphical data. In some embodiments, the sampling resolution includes a number of pixels or voxels that is utilized to visualize a respective graphical data 108 (e.g., visualize a respective 2D image). In some embodiments, the sampling resolution is defined by a processed utilized to capture the respective graphical data 108. [00233] In some embodiments, the first plurality of characteristics includes a first modality (e.g., second characteristic 112-2 of Figure 2A) associated with a capture of the first plurality of graphical data 108-1. For instance, in some embodiments, each respective graphical data 108 is associated with a corresponding modality (e.g., the first modality in a plurality of modalities, the first modality in one or more modalities, etc.). Each respective modality is associated with a unique process and/or technique utilized for capturing the respective graphical data 108. As a non-limiting example, in some embodiments, a first modality is associated with the respective graphical data 108 captured by a single-photo emission computed tomography (SPECT) process and a second modality is associated with the respective graphical data 108 captured by a magnetic resonance imaging (MRI) process. However, the present disclosure is not limited thereto. [00234] Accordingly, in some embodiments, a respective characteristic in the first plurality of characteristics describes a fine-grain aspect of the first plurality of graphical data, such as a particular aspect of a first digital image in a plurality of digital images associated with the DB1/ 147040794.2 56 Attorney Ref. No.: 115834-5037-WO first plurality of graphical data. Moreover, in some embodiments, the respective characteristic in the first plurality of characteristics describes a course-grain aspect of the first plurality of graphical data, such as a particular aspect of the plurality of digital images associated with the first plurality of graphical data. In this way, in some such embodiments, the first plurality of characteristics allow for tuning encoding of a respective graphical data based on either course-grain or fine-grain characteristics associated with the respective graphical data. However, the present disclosure is not limited thereto. [00235] Block 406. Referring to block 406, in some embodiments, the first plurality of graphical data 108-1 includes one or more images taken of a first biological sample. In some embodiments, a respective biological sample is a living or non-living cellular organism. For instance, in some embodiments, the respective biological sample is a living or non-living human, such as a male or female human of any stage (e.g., a man, a woman, a child, etc.). In some embodiments, the respective biological sample includes a 3D body (e.g., a human subject, a flora of a region of interest, an inanimate object, etc.). [00236] In some embodiments, the first plurality of graphical data 108-1 represents a first biological sample during a first epoch. In some embodiments, the second one or more graphical data represents a second biological sample during a second epoch. As a non- limiting example, the first biological sample includes a first 3D body at a first spacetime and the second biological sample includes a second 3D body, different from the first 3D body, at a second spacetime different from the first spacetime. For instance, in some embodiments, the first biological sample is a first chest X-Ray of a first human subject captured on December 1, 2000, and the second biological sample is a second chest X-Ray of a second human captured on January 1, 2023. However, the present disclosure is not limited thereto. [00237] Block 408. Referring to block 408, in some embodiments, the first plurality of graphical data 108-1 includes 2D graphical data, 3D graphical data, 4D graphical data, or a combination thereof. In some embodiments, the first plurality of graphical data includes p- dimensional data, in which p is an integer greater than or equal to two (e.g., four, five, six, fifteen, etc.). For instance, in some embodiments, the 2D graphical data includes a corresponding plurality of pixels that collectively depict one or more digital images. However, the present disclosure is not limited thereto. In some embodiments, the 3D graphical data includes a corresponding plurality of volumetric pixels (voxels) that collectively depict one or more volumetric images. However, the present disclosure is not limited thereto. In some embodiments, a digital asset associated the first plurality of DB1/ 147040794.2 57 Attorney Ref. No.: 115834-5037-WO graphical data 108-1 is two-dimensional (2D) (e.g., defined by x and y spatial coordinates), 3D (e.g., defined by x, y, and z spatial coordinates), four-dimensional (4D) (e.g., defined by x, y, and z spatial coordinates including a temporal coordinate), or n-dimensional (n-D). [00238] In some embodiments, a respective one or more graphical data includes a plurality of space-time characteristics 112, such as one or more characteristics 112 of the environment (e.g., before and/or when capturing a digital asset). In some embodiments, the plurality of space-time characteristics provide a simultaneous localization and mapping (SLAM) of a trajectory. In some embodiments, the plurality of space-time characteristics is used by one or more computational models 118 to generate spatially registered base maps of a 3D body. [00239] By way of example, consider a plurality of dimensions of a n-D graphical data defined by one or more spatial coordinates, one or more temporal coordinates, one or more spectral signatures, one or more physical light characteristics, one or more sampling resolutions, one or more performance thresholds provided by a first model 118-1, one or more training resolutions, or a combination thereof. [00240] Block 410. Referring to block 410, in some embodiments, the first plurality of graphical data 108-1 includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. For instance, in some embodiments, the one or more digital images (e.g., images) include one or more photographs, one or more computer-aided drawings (CAD), one or more maps (e.g., geographical maps, topographical maps, etc.), one or more charts, or a combination thereof. As yet another non-limiting example, in some embodiments, the one or more 3D maps include one or more CT maps, one or more MRI maps, one or more PET maps, or the like, which provide 3D graphical data of an internal region (e.g., internal organs, internal tissue, internal bones) of a biological sample. In some embodiments, the one or more digital videos include one or more animations Accordingly, the method 400 allows for optimizing decoding of a variety of file types utilized to retain a respective plurality of graphical data 108 in 2D and/or 3D forms. [00241] Block 412. Referring to block 412, in some embodiments, the first plurality of graphical data includes 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images. In some embodiments, the first plurality of graphical data 108-1 includes at least 50 digital images, at DB1/ 147040794.2 58 Attorney Ref. No.: 115834-5037-WO least 100 digital images, at least 1,000 digital images, at least 10,000 digital images, or at least 100,000 digital images. In some embodiments, the first plurality of graphical data 108-1 includes at most 50 digital images, at most 100 digital images, at most 1,000 digital images, at most 10,000 digital images, or at most 100,000 digital images. [00242] In some embodiments, the second one or more graphical data 108-2 includes 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images. In some embodiments, the second one or more graphical data 108-1 includes at least 50 digital images, at least 100 digital images, at least 1,000 digital images, at least 10,000 digital images, or at least 100,000 digital images. In some embodiments, the second one or more graphical data 108-1 includes at most 50 digital images, at most 100 digital images, at most 1,000 digital images, at most 10,000 digital images, or at most 100,000 digital images. [00243] Block 414. Referring to block 414, in some embodiments, the sampling resolution of the first plurality of graphical data 108-1 includes the minimum resolution associated with the first plurality of graphical data. For instance, in some embodiments, a respective model 118 imposes a condition (e.g., a first threshold performance criterion) that a native resolution of an initial terminal scan in a plurality of sequence scans must exceed the minimum resolution. In some embodiments, the minimum resolution is a predetermined resolution, such as a first predetermined resolution required for a respective graphical data to be utilized by the respective model 118 or a second predetermined resolution defined, at least in part, by an end-use of the computer system or a capture method for the graphical data. For instance, in some embodiments, the minimum resolution for 2D graphical data 108 is 8x8, 16x16, 32x32, 64x64, 128x128, or 256x256, which is required to obtain sufficient information from the first plurality of graphical data 108-1. In some embodiments, the minimum resolution for 3D graphical data 108 is 8x8x8, 16x16x16, 32x32x32, 64x64x64, 128x128x128, or 256x256x256. In some embodiments, the minimum resolution is defined by one dimension (e.g., a minimum X resolution, a minimum Y resolution, or a minimum Z resolution), by two dimensions (e.g., a minimum X resolution and a minimum Y resolution, the minimum Y resolution and a minimum Z resolution, etc.), or by three dimensions (a minimum X resolution, a minimum Y resolution, and a minimum Z resolution). However, the present disclosure is not limited thereto. [00244] Block 416. Referring to block 416, in some embodiments, the first modality includes a computer tomography (CT) modality (e.g., an Abdomen CT modality, an angiography CT DB1/ 147040794.2 59 Attorney Ref. No.: 115834-5037-WO modality, an arthrography CT modality, a bone CT modality, a brain CT modality, a head CT modality, a chest CT modality, a lung CT modality, a neck CT modality, a pelvis CT modality, a renal stone CT modality, a sinus CT modality, a spine CT modality, etc.), a digital pathology modality, a magnetic resonance imaging (MRI) modality (e.g., a functional MRI modality, a breast MRI modality, a magnetic resonance angiography MRI modality, a magnetic resonance venographic MRI modality, a cardiac MRI modality, etc.), a positron emission tomography (PET) modality (e.g., a PET f-18 FDG modality, a cardiac PET modality, a PET/CT modality, etc.), a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof. For instance, in some embodiments, the MRI modality includes spin-echo imaging, inversion recovering imaging, echo planar imaging, gradient echo imaging, and/or the like. In some embodiments, the sonography modality includes ultrasound imaging, doppler imaging, and/or the like. [00245] In some embodiments, a respective modality is specific to both a method of capture and a subject matter association with a region of interest (ROI) of a corresponding graphical data, such as a first abdominal CT modality, a second brain MRI, a third chest X-ray modality, a fourth cardiac abdomen ultrasound modality, a fifth fluorescence microscopy modality, a sixth retinal fundoscopy modality, and a seventh statistical graphics modality, or a combination thereof. However, the present disclosure is not limited thereto. [00246] Block 418. Referring to block 418 of Figure 4B, in some embodiments, the method 400 includes using the first plurality of graphical data 108-1 to train a feature extraction model (e.g., first model 118-1 of Figure 2B). In some embodiments, from this training, the method 400 obtains a trained feature extraction model (e.g., second model 118-2 of Figure 2B). [00247] In some embodiments, the training the feature extraction model further obtains performance data. For instance, in some embodiments, the performance data includes a plurality of training resolutions. In some embodiments, each respective training resolution in the plurality of training resolutions is less than the sampling resolution, which allows for each respective training resolution to be associated with a reduced data size format. Moreover, in some embodiments, each respective training resolution is associated with a corresponding threshold performance for the trained feature extraction model 118 in a plurality of threshold performances. For instance, in some embodiments, a first training resolution (e.g., a 32x32 resolution) is associated with a first threshold performance when using the first training resolution with a first model 118-1, and a second training resolution (e.g., a 64x32 resolution) DB1/ 147040794.2 60 Attorney Ref. No.: 115834-5037-WO is associated with a second threshold performance greater than the first threshold performance when using the second training resolution with the first model 118-1. However, the present disclosure is not limited thereto. [00248] Block 420. Referring to block 420, in some embodiments, the feature extraction model 118 includes a segmentation model (e.g., first model 118-1 of Figure 2B) and/or a classification model (e.g., second model 118-2 of Figure 2B). As a non-limiting example, in some embodiments, the one or more feature extraction models 118 includes one or more convolutional neural networks (CNN) models 118, one or more recurrent neutral networks (RNN) models 118, one or more support vector machines (SVM) models 118, one or more decision tree models 118, one or more clustering models 118, one or more random forest models 118, one or more deep belief network models 118, or a combination thereof. For instance, in some embodiments, a respective classification model 118 is utilized to assign a label to an input feature associated with a respective plurality of graphical data 108 based on a characteristic of the respective plurality of graphical data 108. In some embodiments, a respective segmentation model 118 partitions the respective plurality of graphical data 108 based on distinct features or regions present in the respective plurality of graphical data 108. [00249] Block 422. Referring to block 422, in some embodiments, the feature extraction model includes a neutral network model, a support machine mode, a Naïve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph- based model, or a combination thereof. [00250] Block 424. Referring to block 424, in some embodiments, the first plurality of characteristics 110 includes one or more ground truth labels (e.g., fourth characteristic 112-4 of Figure 2A). In some embodiments, each ground truth label in the one or more ground truth labels provides an indication of a presence or an absence of a respective feature, such as an abnormality. However, the present disclosure is not limited thereto. For instance, in some embodiments, a respective ground truth label is obtained from a measurement associated with the graphical data 108, such as by a first model 118 utilized to obtain the one or more ground truth labels. In some embodiments, the respective ground truth label is obtained from a source of the respective graphical data 108, such as an administrator of a second computer system (e.g., client device 300 of Figure 1, client device 300 of Figure 3, etc.) or the first computer system (e.g., graphical data system 200 of Figure 1, graphical data system 200 of Figures 2A and 2B, etc.). Accordingly, each respective ground truth label provides an DB1/ 147040794.2 61 Attorney Ref. No.: 115834-5037-WO indication of a known corresponding outcome, output, or conclusion based on an associated input of the respective graphical data 108. [00251] Block 426. Referring to block 426, in some embodiments, the first plurality of graphical data 108-1 is used to train the feature extraction model 118 against the one or more ground truth labels (e.g., first characteristic 112-1 of Figure 2A). [00252] Block 428. Referring to block 428 of Figure 4C, in some embodiments, the method 400 includes receiving a request to evaluate a second one or more graphical data (e.g., second graphical data 108-2 of Figure 2A). In some embodiments, the request to evaluate the second one or more graphical data 108 is a request to perform the evaluation at a second computer system. Accordingly, the method 400 allows for optimizing decoding of the second one or more graphical data 108-2 at the second computer system based on using the first plurality of graphical data 108 with the feature extraction model 118 at the first computer system. However, the present disclosure is not limited thereto. In some embodiments, the request to evaluate the second one or more graphical data 108 is a request to perform the evaluation at the first computer system. [00253] In some embodiments, the second one or more graphical data 108-2 is defined by a second plurality of characteristics (e.g., second characteristics 110-2 of Figure 2A. In some embodiments, the second plurality of characteristics is different than the first plurality of characteristics of the first plurality of graphical data 108-1. However, the present disclosure is not limited thereto. In some embodiments, the second plurality of characteristics 110-2 includes one or more characteristics including the first modality associated with a capture of the second one or more graphical data 108-1. Accordingly, in some embodiments, the process of capturing the first plurality of graphical data 108-1 and the second one or more graphical data 108-2 are the same (e.g., both the first plurality of graphical data 108-1 and the second one or more graphical data 108-2 were captured by an MRI process, both the first plurality of graphical data 108-1 and the second one or more graphical data 108-2 were captured by CT process, etc.). [00254] In some embodiments, the second plurality of characteristics 110-2 includes one or more characteristics including a desired threshold performance. In some embodiments, the desired performance includes a threshold accuracy and/or a threshold precision that must be satisfied by a respective training resolution when applying the second one or more graphical data 108-2 to a respective feature extraction model 118. In some embodiments, the desired DB1/ 147040794.2 62 Attorney Ref. No.: 115834-5037-WO threshold performance is based on a characteristic of the second one or more graphical data 108-2, the first plurality of graphical data 108-1, a parameter 120 of the respective feature extraction model 118, or a combination thereof. However, the present disclosure is not limited thereto. [00255] In some embodiments, the desired threshold performance is based on a first evaluation of a first training resolution. In some embodiments, the desired threshold performance is based, at least in part, on using the sampling resolution with a respective feature extraction model 118 in one or more feature extraction models at the first computer system. For instance, in some embodiments, the desired threshold performance defines a quality metric that is a minimum ability to extract information from any resolution (e.g., any scan in a plurality of sequence scans) from an encrypted byte stream 114 that maintains full information efficiency. [00256] Block 430. Referring to block 430, in some embodiments, the second one or more graphical data 108-2 includes one or more images taken of a second biological sample different than the first biological sample. As such, in some such embodiments, different biological samples are utilized to create a respective graphical data 108. For instance, in some embodiments, the first biological sample is a first human subject and the second biological sample is a second human. Accordingly, the method 400 allows for training a model 118 using a first biological sample and evaluating the second biological sample based on the training with the first biological sample. However, the present disclosure is not limited thereto. In some embodiments, the first biological sample is an initial sample used to validate and/or train a respective model 118, such as in order to identify one or more patterns associated with the first biological sample, and the second biological sample is a real-world dataset. As a further non-limiting example, consider a human 3D body with each sub-body in one or more bodies defined at a point of articulation of the human 3D body. In some embodiments, the systems and methods of the present disclosure obtain a first plurality of graphical data 108-1 associated with a first sub-body in the one or more sub-bodies and a second one or more graphical data 108-2 associated with a second sub-body in the one or more sub-bodies (e.g., first graphical data is of a face, second digital data is of a foot, etc.). In some such embodiments, a third plurality of graphical data 108-3 is generated based on the first and second pluralities of graphical data 108, in which the third plurality of graphical is associated with a combination of the first and second pluralities of graphical 108. However, the present disclosure is not limited thereto. DB1/ 147040794.2 63 Attorney Ref. No.: 115834-5037-WO [00257] Block 432. Referring to block 432, in some embodiments, the second one or more graphical data 108-2 includes 3D graphical data. Moreover, each scan in the plurality of sequence scans of the 3D graphical data includes a respective 2D layer of the 3D graphical data. For instance, in some embodiments, each 2D layer is formed by applying a splicing technique to the 3D graphical data. As a non-limiting example, in some embodiments, for a respective point cloud graphical data 108, a first layer includes 1 x 102 points, a second layer includes 1 x 103 points, a third layer includes 1 x 105 points, and the like. However, the present disclosure is not limited thereto. [00258] Block 434. Referring to block 434, in some embodiments, the second one or more graphical data includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. For instance, as a non-limiting example, in some embodiments, the method 400 includes point cloud pre-processing, spatial clustering segmentation based on K-dimensional tree, acquisition of structural parameters, calculation of volume based on multiple regression analysis, and the like. [00259] Block 436. Referring to block 436, in some embodiments, the second plurality of characteristics further includes a respective capacity of the communication network 186. In some embodiments, a first epoch is required to communicate a respective plurality of graphical data 108 via a communication network 186, such that the respective capacity of the communication network 186 limits an ability to communicate a respective plurality of graphical data 108 through an encoded byte stream. For instance, in some embodiments, the respective capacity of the communication network 186 includes a bandwidth of the communication network 186, which defines an amount of data that is communicated per unit time (e.g., 100 megabyte (Mb) per second, etc.). In some embodiments, the bandwidth of the communication network 186 changes with time. Accordingly, in some such embodiments, in order to prevent a prolonged period of time for communicating the encoded byte stream, the desired threshold performance dynamically modifies to similar to the change in the bandwidth of the communication network 186. However, the present disclosure is not limited thereto. For instance, in some embodiments, the respective capacity of the communication network 186 includes a latency of the communication network 186, which defines an epoch required to communicate a respective data packet (e.g., a respective scan in a plurality of sequence scans of the encoded byte stream). In some embodiments, the DB1/ 147040794.2 64 Attorney Ref. No.: 115834-5037-WO respective capacity of the communication network 186 includes a data capacity of the communication network 186, which defines the maximum size of data communicated during the epoch. [00260] Block 438. Referring to block 438, in some embodiments, the second plurality of characteristics further includes a respective capacity of the first computer system (e.g., graphical data system 200 of Figures 2A and 2B, client device 300-R of Figure 1, client device 300 of Figure 3, etc.). For instance, in some embodiments, the respective capacity of the first computer system is a hardware limitation associated with utilizing a respective model 118 at the first computer system. As a non-limiting example, in some embodiments, the respective capacity of the first computer system includes a processing power of the first computer system, a memory requirement of first computer system, a storage requirement of first computer system, a heat dissipation requirement of the first computer system, a power consumption requirement of first computer system, or a combination thereof. However, the present disclosure is not limited thereto. [00261] In some embodiments, the respective graphical data includes one or more images of the biological sample on the substrate is obtained. Each of the one or more images includes a corresponding plurality of pixels in the form of an array of pixel values. In some embodiments the array of pixel values includes at least a least 100, 10,000, 100,000, 1 x 106, 2 x 106, 3 x 106, 5 x 106, 8 x 106, 10 x 106, or 15 x 106 pixel values. [00262] In some embodiments, the single intensity threshold is determined using Otsu’s method, where the first heuristic classifier identifies a threshold that minimizes intra-class variance or equivalently maximizes inter-class variance. In some such embodiments, Otsu’s method uses a discriminative analysis that determines an intensity threshold such that binned subsets of pixels in the plurality of pixels are as clearly separated as possible. Each respective pixel in the plurality of pixels is binned or grouped into different classes depending on whether the respective intensity value of the respective pixel falls over or under the intensity threshold. For example, in some embodiments, bins are represented as a histogram, and the intensity threshold is identified such that the histogram can be assumed to have a bimodal distribution (e.g., two peaks) and a clear distinction between peaks (e.g., valley). [00263] In some such embodiments, the plurality of pixels in the obtained image is filtered such that pixels including a pixel intensity above the intensity threshold are considered to be foreground and are converted to white (e.g., uint8 value of 1), while pixels including a pixel DB1/ 147040794.2 65 Attorney Ref. No.: 115834-5037-WO intensity below the intensity threshold are considered to be background and are converted to black (e.g., uint8 value of 0). In some embodiments, an outcome of a heuristic classifier using Otsu’s method is illustrated (e.g., using display 182 of Figure 2A), which depicts a thresholded image (e.g., a mask or a layer) after conversion of the acquired image, where each pixel in the plurality of pixels is represented as either a white or a black pixel. Here, Otsu’s method is an example of a binarization method using global thresholding. In some embodiments, Otsu’s method is robust when the variances of the two classes (e.g., foreground and background) are smaller than the mean variance over the obtained image as a whole. [00264] In some embodiments, the first heuristic classifier is a binarization method other than Otsu’s method. In some such embodiments, the first heuristic classifier is a global thresholding method other than Otsu’s method or an optimization-based binarization method. In some such embodiments, a global thresholding method is performed by determining the intensity threshold value manually (e.g., via default or user input). For example, an intensity threshold can be determined at the middle value of the grey-scale range (e.g., 128 between 0- 255). [00265] In some embodiments, the intensity threshold value is determined automatically using a histogram of grey-scale pixel values (e.g., using the mode method and/or P-tile method). For example, using the mode method, a histogram of grey-scale pixel values can include a plurality of bins (e.g., up to 256 bins for each possible grey-scale pixel value 0- 255), and each respective bin is populated with each respective pixel having the respective grey-scale pixel value. In some embodiments, the plurality of bins has a bimodal distribution and the intensity threshold value is the grey-scale pixel value at which the histogram reaches a minimum (e.g., at the bottom of the valley). Using the P-tile method, each respective bin in a histogram of grey-scale pixel values is populated with each respective pixel having the respective grey-scale pixel value, and a cumulative tally of pixels is calculated for each bin from the highest grey-scale pixel value to the lowest grey-scale pixel value. Given a pre- defined number of pixels P above the intensity threshold value, the threshold value is determined at the bin value at which the cumulative sum of pixels exceed P. [00266] In some embodiments, an intensity threshold value is determined by estimating the level of background noise (e.g., in imaging devices including but not limited to fluorescence microscopy). Background noise can be determined using control samples and/or unstained samples during normalization and pre-processing. DB1/ 147040794.2 66 Attorney Ref. No.: 115834-5037-WO [00267] In some embodiments, such as when using optimization-based binarization, the assignment of a respective pixel to one of two classes (e.g., conversion to either black or white) is determined by calculating the relative closeness of the converted pixel value to the original pixel value, as well as the relative closeness of the converted pixel value of the respective pixel to the converted pixel values of neighboring pixels (e.g., using a Markov random field). Optimization-based methods thus include a smoothing filter that reduces the appearance of small punctate regions of black and/or white and ensures that local neighborhoods exhibit relatively congruent results after binarization. See, Uchida, 2013, “Image processing and recognition for biological images,” Develop. Growth Differ.55, 523- 549, doi:10.1111/dgd.12054, which is hereby incorporated herein by reference in its entirety. [00268] In some embodiments, the first or second heuristic classifier include a smoothing method to minimize or reduce noise between respective pixels in a local neighborhood by filtering for differences in pixel intensity values. In some embodiments, smoothing is performed in a plurality of pixels in grey-scale space. In some embodiments, applicable smoothing methods include, but are not limited to, blurring filters, median filters, and/or bilateral filters. For example, in some embodiments, a blurring filter minimizes differences within a local neighborhood by replacing the pixel intensity values at each respective pixel with the average intensity values of the local neighborhood around the respective pixel. In some embodiments, a median filter utilizes a similar method, but replaces the pixel intensity values at each respective pixel with the median pixel values of the local neighborhood around the respective pixel. Whereas, in some embodiments, blurring filters and median filters cause image masks to exhibit “fuzzy” edges, in some alternative embodiments, a bilateral filter preserves edges by determining the difference in intensity between pixels in a local neighborhood and reducing the smoothing effect in regions where a large difference is observed (e.g., at an edge). See, Uchida, 2013, “Image processing and recognition for biological images,” Develop. Growth Differ.55, 523-549, doi:10.1111/dgd.12054, which is hereby incorporated herein by reference in its entirety. [00269] In some embodiments, a global thresholding method is further applied to an image mask including the outcome of a local intensity gradient filter represented as an array (e.g., a matrix) of grey-scale pixel values. In some such embodiments, the local intensity gradient array is binarized into two classes using Otsu’s method, such that each pixel in the plurality of pixels is converted to a white or a black pixel (e.g., having pixel value of 1 or 0, respectively), representing foreground or background, respectively. In some embodiments, DB1/ 147040794.2 67 Attorney Ref. No.: 115834-5037-WO Otsu’s method is applied to a local intensity gradient filter from an obtained image, such that binarization is applied to regions of high and low granularity rather than regions of high and low pixel intensity. This provides an alternative method for classifying foreground and background regions over global thresholding methods. [00270] In some embodiments, a second heuristic classifier is a two-dimensional Otsu’s method, which, in some instances, provides better image segmentation for images with high background noise. In the two-dimensional Otsu’s method, the grey-scale intensity value of a respective pixel is compared with the average intensity of a local neighborhood. Rather than determining a global intensity threshold over the entire image, an average intensity value is calculated for a local neighborhood within a fixed distance radius around the respective pixel, and each pair of intensity values (e.g., a value averaged over the local neighborhood and a value for the respective pixel) are binned into a discrete number of bins. The number of instances of each pair of average intensity values for the local neighborhood and for the respective pixel, divided by the number of pixels in the plurality of pixels, determines a joint probability mass function in a 2-dimensional histogram. In some embodiments, the local neighborhood is defined by a disk including a radius of fixed length (e.g., 1, 2, 3, 4, 5, 6, 7, 8, 9, or 10 pixels, between 10 and 50 pixels, between 50 and 100 pixels, between 100 and 200 pixels, or more than 200 pixels). [00271] Image segmentation. In some embodiments, the method for tissue classification further includes applying the aggregated score and intensity of each respective pixel in the plurality of pixels to a segmentation algorithm, such as graph cut, to independently assign a probability to each respective pixel in the plurality of pixels of being tissue sample or background. [00272] In some embodiments, graph cut performs segmentation of a monochrome image based on an initial trimap T = {TB, TU, TF}, where TB indicates background regions, TF indicates foreground regions, and TU indicates unknown regions. The image is represented as an array z = (z1, … , zn, …, zN) including grey-scale pixel values for a respective pixel n in a plurality of N pixels. As in Bayes matting models, the graph cut segmentation algorithm attempts to compute the alpha values for TU given input regions for TB and TF, by creating an alpha-matte that reflects the proportion of foreground and background for each respective pixel in a plurality of pixels as an alpha value between 0 and 1, where 0 indicates background and 1 indicates foreground. In some embodiments, an alpha value is computed by transforming a grey-scale pixel value (e.g., for an 8-bit single-channel pixel value between 0 DB1/ 147040794.2 68 Attorney Ref. No.: 115834-5037-WO and 255, the pixel value is divided by 255). Graph cut is an optimization-based binarization technique as described above, which uses polynomial-order computations to achieve robust segmentation even when foreground and background pixel intensities are poorly segregated. See, Rother et al., 2004, “‘GrabCut’ – Interactive Foreground Extraction using Iterated Graph Cuts,” ACM Transactions on Graphics.23(3):309-314, doi:10.1145/1186562.1015720, which is hereby incorporated herein by reference in its entirety. See also, Boykov and Jolly, 2001, “Interactive graph cuts for optimal boundary and region segmentation of objects in N-D images,” Proc. IEEE Int. Conf. on Computer Vision, CD-ROM, and Greig et al., 1989, “Exact MAP estimation for binary images,” J. Roy. Stat. Soc. B.51, 271-279, for details on graph cut segmentation algorithms; and Chuang et al., 2001, “A Bayesian approach to digital matting,” Proc. IEEE Conf. Computer Vision and Pattern Recog., CD-ROM, for details on Bayes matting models and alpha-mattes, each of which is hereby incorporated herein by reference in its entirety. [00273] In some embodiments, the trimap is user specified. In some embodiments, the trimap is initialized using the plurality of heuristic classifiers as an initial tissue detection function. In some such embodiments, the set of classes including obvious first class, likely first class, likely second class, and obvious second class are provided to the graph cut segmentation algorithm as a trimap including TF = {obvious first class} (e.g., obvious foreground), TB = {obvious second class} (e.g., obvious background), and TU = {likely first class, likely second class} (e.g., concatenation of likely foreground and likely background). In some embodiments, the TF = {obvious first class, probable first class} (e.g., obvious foreground and probable foreground), TB = {obvious second class, probable second class} (e.g., obvious background and probable background), and TU is any unclassified pixels in the plurality of pixels in the obtained image. In some embodiments, the set of classes is provided to the graph cut segmentation algorithm using an alternate trimap that is a combination or substitution of the above implementations that will be apparent to one skilled in the art. [00274] In some embodiments, the segmentation algorithm is a GrabCut segmentation algorithm. The GrabCut segmentation algorithm is based on a graph cut segmentation algorithm, but includes an iterative estimation and incomplete labelling function that limits the level of user input required and utilizes an alpha computation method used for border matting to reduce visible artefacts. Furthermore, GrabCut uses a soft segmentation approach rather than a hard segmentation approach. Unlike graph cut segmentation algorithms, GrabCut uses Gaussian Mixture Models (GMMs) instead of histograms of labelled trimap DB1/ 147040794.2 69 Attorney Ref. No.: 115834-5037-WO pixels, where a GMM for a background and a GMM for a foreground are full-covariance Gaussian mixtures with K components. To make the GMM a tractable computation, a unique GMM component is assigned to each pixel in the plurality of pixels from either the background or the foreground model (e.g., 0 or 1). See, Rother et al., 2004, “‘GrabCut’ – Interactive Foreground Extraction using Iterated Graph Cuts,” ACM Transactions on Graphics.23(3):309-314, doi:10.1145/1186562.1015720, which is hereby incorporated herein by reference in its entirety. [00275] In some embodiments, the GrabCut segmentation algorithm can operate either on a multi-spectral, multi-channel image (e.g., a 3-channel image) or on a single-channel image. In some embodiments, a grey-scale image is provided to the segmentation algorithm. In some embodiments, a grey-scale image is first converted to a multi-spectral, multi-channel image (e.g., RGB, HSV, CMYK) prior to input into the segmentation algorithm. In some embodiments, a multi-spectral, multi-channel color image is applied directly to the segmentation algorithm. [00276] In some embodiments, the GrabCut segmentation algorithm is applied to the image as a convolution method, such that local neighborhoods are first assigned to a classification (e.g., foreground or background) and assignations are then applied to a larger area. In some embodiments, an image including a plurality of pixels is provided to the GrabCut algorithm as a color image, using the initialization labels obtained from the plurality of heuristic classifiers, and the binary classification output of the GrabCut algorithm is used for downstream spatial analysis (e.g., on barcoded capture spots). In some embodiments, the plurality of pixels assigned with a greater probability of tissue or background is used to generate a separate construct (e.g., a matrix, array, list or vector) indicating the positions of tissue and the positions of background in the plurality of pixels. The GrabCut segmentation algorithm performs binary identification of tissue and background, which is evident from the clear isolation of the tissue section overlay from the background regions. [00277] In some embodiments, the aggregated score and intensity of each respective pixel in the plurality of pixels is applied to a segmentation algorithm other than a graph cut segmentation algorithm or a GrabCut segmentation algorithm, including but not limited to, Magic Wand, Intelligent Scissors, Bayes Matting, Knockout 2, level sets, binarization, background subtraction, watershed method, region growing, clustering, active contour model (e.g., SNAKES), template matching and recognition-based method, Markov random field. In some embodiments, the aggregated score and intensity of each respective pixel in the DB1/ 147040794.2 70 Attorney Ref. No.: 115834-5037-WO plurality of pixels is applied to a feature extraction algorithm (e.g., intuition and/or heuristics, gradient analysis, frequency analysis, histogram analysis, linear projection to a trained low- dimensional subspace, structural representation, and/or comparison with another image). In some embodiments, the aggregated score and intensity of each respective pixel in the plurality of pixels is applied to a pattern classification method including but not limited to nearest neighbor classifiers, discriminant function methods (e.g., Bayesian classifier, linear classifier, piecewise linear classifier, quadratic classifier, support vector machine, multilayer perception/neural network, voting), and/or classifier ensemble methods (e.g., boosting, decision tree/random forest). See, Rother et al., 2004, “‘GrabCut’ – Interactive Foreground Extraction using Iterated Graph Cuts,” ACM Transactions on Graphics.23(3):309-314, doi:10.1145/1186562.1015720, and, Uchida, 2013, “Image processing and recognition for biological images,” Develop. Growth Differ.55, 523-549, doi:10.1111/dgd.12054, each of which is hereby incorporated herein by reference in its entirety. [00278] In some embodiments, the plurality of pixels is in, or is converted to, grey-scale space by obtaining the image in grey-scale (e.g., a single-channel image), or by obtaining the image in color (e.g., a multi-channel image) and converting the image to grey-scale after the obtaining and prior to the running of the heuristic classifiers. In some embodiments, each respective pixel in the plurality of pixels in grey-scale space has an integer value between 0 and 255 (e.g., 8-bit unsigned integer value or “uint8”). In some embodiments, the integer value for each respective pixel in the plurality of pixels of the image in grey-scale space is transformed using e.g., addition, subtraction, multiplication, or division by a value N, where N is any real number. For example, in some embodiments, each respective pixel in the plurality of pixels in grey-scale space has an integer value between 0 and 255, and each integer value for each respective pixel is divided by 255, thus providing integer values between 0 and 1. In some embodiments, the plurality of pixels of the image is in grey-scale space and is transformed using contrast enhancement or tone curve alignment. In some embodiments, the running of the plurality of heuristic classifiers on the plurality of pixels includes rotating, transforming, resizing, or cropping the obtained image in grey-scale space. [00279] Block 440. Referring to block 440, in some embodiments, the desired threshold performance includes a threshold inter-image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation. In some embodiments, the threshold inter- image correlation includes a first requirement for a degree of similarity between a first graphical data and a second graphical data (e.g., a first degree of similarity between a first DB1/ 147040794.2 71 Attorney Ref. No.: 115834-5037-WO image of the second one or more graphical data and a second image of the second one or more graphical data, a first degree of similarity between the first image of the second one or more graphical data and a first image of the first plurality of graphical data, etc.). In some embodiments, the threshold intra-image correlation includes a second requirement for a degree of similarity within a respective graphical data. As a non-limiting example, in some embodiments, the degree of similarity is between two or more regions of a respective 2D or 3D image and/or between two or more features of the respective 2D or 3D image. For instance, a relatively larger degree of similarity means that the two or more regions of the respective 2D or 3D image and/or the two or more features of the respective 2D or 3D image have a high similarity. Furthermore, in some embodiments, the threshold inter-and-intra- image correlation includes a requirement for a degree of similarity between two or more regions and/or two or more features of a first 2D or 3D image and of between the first 2D or 3D image and a second 2D or 3D image, which allows for the method 400 to identify patterns both within a respective 3D or 3D image or between 2D and 3D images of graphical data 108. [00280] Block 442. Referring to block 442 of Figure 4D, in some embodiments, the method 400 further includes encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data 108-2. From this encoding, the method 400 forms an encoded byte stream (e.g., first encoded byte stream 114-1 of Figure 2A). In some embodiments, the encoded byte stream includes a plurality of sequence scans through the sampling resolution, which allows an end-user to view, at a highest resolution, the sampling resolution of the second one or more graphical data 108-2. In some embodiments, each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order. In some embodiments, each rank order defines a sequence of respective sequence scans in the plurality of sequence scans. For instance, in some embodiments, the rank order defines a structured organized set that assigns each unique rank in the rank order to a position in the rank order. In some embodiments, the rank order is based on a value of a corresponding resolution. Furthermore, in some embodiments, each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. DB1/ 147040794.2 72 Attorney Ref. No.: 115834-5037-WO [00281] Block 444. Referring to block 444, in some embodiments, the first codec 124-1 includes a predictive codec 124, an embedded codec 124, a sub-band codec 124, a block- based codec 124, a layered codec 124, a lossless codec 124, a lossy codec 124, or a combination thereof. However, the present disclosure is not limited thereto. [00282] For instance, in some embodiments, the predictive codec 124 includes one or more instructions that predict content (e.g., subject matter) for a portion of a respective plurality of graphical data 108 based on information learned from the respective plurality of graphical data 108. As a non-limiting example, in some embodiments, the predictive codec the predictive codec 124 includes one or more instructions that predict a feature of second 2D image in a second one or more graphical data 108-2 based on a first 2D image in a first plurality of graphical data 108-1. However, the present disclosure is not limited thereto. [00283] In some embodiments, the embedded codec 124 includes one or more instructions that is integrated (e.g., embedded, included) within a corresponding storage format (e.g., a file type, a container, a mountable image, etc.). [00284] In some embodiments, the sub-band codec 124 includes one or more instructions for segmenting a respective plurality of graphical data 108 into two or more bands (e.g., two or more discrete frequency bands), in which each respective band in the two or more bands is processed individually by a corresponding encoder 126 and/or a corresponding decoder 128 of the sub-band codec 124. [00285] In some embodiments, the block-based codec 124, the sub-band codec 124 includes one or more instructions for segmenting a respective plurality of 2D graphical data 108, particularly video 2D graphical data, into two or more sub-blocks), in which each respective band in the two or more bands is processed individually by a corresponding encoder 126 and/or a corresponding decoder 128 of the block-based codec 124. [00286] In some embodiments, the layered codec 124 includes one or more instructions for layering a respective plurality of graphical 108 into two or more discrete layers that are collectively stored in a corresponding storage format (e.g., a file type, a container, a mountable image, etc.). [00287] In some embodiments, the lossless codec 124 utilizing a corresponding encoder 126 and a corresponding decoder 128 that communicates a respective plurality of graphical data 108 through an encoded byte stream without omitting a portion of information, which allows for maintaining high quality information within the encoded byte stream. DB1/ 147040794.2 73 Attorney Ref. No.: 115834-5037-WO [00288] In some embodiments, the lossy codec 124 utilizing a corresponding encoder 126 and a corresponding decoder 128 that communicates a respective plurality of graphical data 108 through an encoded byte stream by omitting a portion of information, such as a first portion of information that is not requirement to satisfy the threshold performance criterion, which allows for a reduced file size of the encoded byte stream. [00289] Block 446. Referring to block 446, in some embodiments, the first codec 124-1 is selected by the first computer system (e.g., graphical data system 200 of Figure 2A and 2B, client device 300 of Figure 3, etc.). For instance, in some embodiments, the first computer system requires using the second one or more graphical data 108-2 in a first file type compatible with the first codec 124-1. As another non-limiting example, in some embodiments, a first model 118-1 at the first computer system requires receiving a respective plurality of graphical data 108 in accordance with the first codec 124-1. However, the present disclosure is not limited thereto. [00290] Block 448. Referring to block 448, in some embodiments, the plurality of sequence scans includes between 2 scans and 100 scans, between 3 and 70 scans, between 4 and 50 scans, between 5 and 30 scans, between 6 and 20 scans, between 7 and 15 scans, or between 8 and 12 scans. [00291] In some embodiment, the plurality of sequenced scans includes at least 3 scans, at least 4 scans, at least 5 scans, at least 6 scans, at least 7 scans, at least 8 scans, at least 9 scans, at least 10 scans, at least 12 scans, at least 15 scans, at least 20 scans, at least 30 scans, at least 50 scans, or at least 100 scans. [00292] In some embodiment, the plurality of sequenced scans includes at most 3 scans, at most 4 scans, at most 5 scans, at most 6 scans, at most 7 scans, at most 8 scans, at most 9 scans, at most 10 scans, at most 12 scans, at most 15 scans, at most 20 scans, at most 30 scans, at most 50 scans, or at most 100 scans. [00293] Block 450. Referring to block 450, in some embodiments, the plurality of sequence scans includes N scans. In some embodiments, ^^^^ is� ^^^^ ^^^^ ^^^^ ^^^^ 2� ^^^^ ^^^^ ^^^^ ^^^^ ^^^^√2��. In some embodiments, M is a number of digital assets associated with the second one or more graphical data. In some embodiments, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data. Moreover, in some embodiments, ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. DB1/ 147040794.2 74 Attorney Ref. No.: 115834-5037-WO [00294] Block 452. Referring to block 452, in some embodiments, the plurality of sequence scans includes N scans. In some embodiments, ^^^^ is ^^^^ ^^^^ 2 − 1. In some
Figure imgf000077_0001
embodiments, M is a number of digital assets associated with the second one or more graphical data. In some embodiments, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data. Moreover, in some embodiments, ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. Accordingly, in some embodiments, the method 400 allows for extracting information from any power of 2-related resolution from the encoded byte stream while maintaining the information (e.g., lossless information, about lossless information, lossy information) from the second one or more graphical data 108-2. [00295] In some embodiments, the plurality of sequence scans 1010 includes N scans, in which ^^^^ is determined for each graphical data in the one or more graphical data 108 separately prior to forming the encoded byte stream map 600 and/or encoded byte stream 114, which ensured that the corresponding resolution of a sequence scan 1010 remained consistent across medical images with different resolutions as follows ^^^^
Figure imgf000077_0002
where ^^^^ and ^^^^ are a horizontal resolution and a vertical resolution, respectively, of the graphical data 108, and ^^^^ is a lower bound for the resolution of the initial terminal scan. [00296] Block 454. Referring to block 454, in some embodiments, the second plurality of characteristics includes a sampling resolution of the second one or more graphical data. [00297] Moreover, in some embodiments, the plurality of sequence scans includes N scans. In some embodiments, N is based, at least in part, on the sampling resolution of the second one or more graphical data. For instance, in some such embodiments, N is selected based on the sampling resolution of the second one or more graphical data. [00298] Referring to Figure 6, in some embodiments, the second one or more graphical data 108 is progressively encoded using a corresponding encoder 126 of a lossless HTJ2K codec in the form of a plurality of sequence scans that includes N scans. In some embodiments, a respective scan is also known as a respective decomposition or a respective subset of the encoded byte stream. In some embodiments, each respective scan has a native resolution as a factor of 2N of the full graphical data resolution (e.g., resolution associated with a final terminal scan in the plurality of sequence scans, the maximum resolution, etc.). In some embodiments, the native resolution of a respective scan in the plurality of sequence scans DB1/ 147040794.2 75 Attorney Ref. No.: 115834-5037-WO refers to a proportion of pixels of the final terminal scan in the plurality of sequence scans communication prior to any interpolation. In some embodiments, unlike a sequential encoded byte stream of a JPEG codec 124, a HTJ2K encoded byte stream is communicated in the form of the plurality of sequence scans including N scans, as illustrated in Figure 6. For instance, in some embodiments, a second one or more graphical data associated with a first 2D chest x-ray (CXR) modality with a resolution of 256x256 is progressively encoded using a corresponding encoder 126 into an encoded byte stream including 3 scans in a plurality of sequence scans with native resolutions 64x64, 128x128, and 256x256, respectively. Accordingly, the method 400 allows the HTJ2K codec 124 to not just achieve a high compression ratio, but also enables quick access to smaller resolution scans in the plurality of sequence scans of the second one or more graphical data, thus minimizing data duplication. For instance, in Figure 25, a visualization of the second one or more graphical data is formed by using about 25% of the encoded byte stream using the JPEG codec and the HTJ2K codec, respectively. [00299] Block 456. Referring to block 456 of Figure 4E, in some embodiments, the method 400 includes matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data. For instance, in some embodiments, referring briefly to chart 1200 of Figure 11, the method matches the desired threshold performance to the first training resolution in accordance with one or more logical operation based on some Boolean operation. For instance, in some embodiments, a first logical operation describes an “AND” Boolean operation that requires both elements of the first logical operation to be satisfied for a respective threshold to be deemed satisfied. A second logical operation describes an “OR” Boolean operation that requires any one element of the second logical operation to be satisfied for a respective threshold to be deemed satisfied. Moreover, a third logical operation describes an “EXCLUSIVE OR (XOR)” Boolean operation that requires any one element of the third logical operation to be satisfied and no other element satisfied for a respective threshold to be deemed satisfied. A fourth logical operation describes a singular “NOT” Boolean operation that requires absence of an element of the fourth logical operation to be satisfied for a respective threshold to be deemed satisfied. A fifth logical operation describes a plural “NOT” Boolean operation that requires both absence of a first element and presence of a second element of the fifth logical operation to be satisfied for a respective threshold to be deemed satisfied. In some embodiments, a logical operation of a respective model includes a combination of one or more of the above described DB1/ 147040794.2 76 Attorney Ref. No.: 115834-5037-WO logical operations. For instance, in some embodiments, a respective logical operation includes one or more AND, OR, XOR, or NOT operations within the respective logical operation that is utilized to control a subset of a plurality of microfluid devices 300 associated with the device 100. However, the present disclosure is not limited thereto. [00300] In some embodiments, the method matches the desired threshold performance to the first training resolution in accordance with a determination the desired threshold performance is satisfied or not satisfied by taking a cosine similarity measure or dot product of the desired threshold performance against each training resolution in the plurality of training resolutions. [00301] Block 458. Referring to block 458, in some embodiments, the method 400 includes communicating, via a communication network 186, in accordance with the first codec 124, to a second computer system (e.g., client device 300 of Figure 3), the encoded byte stream 114. In some embodiments, the method 400 further includes generating and/or communicating one or more instructions to terminate decoding of the encoded byte stream 114 at the second computer system. In some embodiments, the one or more instructions instruct to terminate decoding when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution. From this communicating of the encoded byte stream 114 and the one or more instructions, the method 400 optimizing decoding of graphical data 108, such as by preventing the second computer system 200 from having to receiving excess scans that exceed the first training resolution. However, the present disclosure is not limited thereto. [00302] Block 460. Referring to block 460, in some embodiments, the first sequence scan includes less than 5% of the encoded byte stream. In some embodiments, the first sequence scan includes between 0.1% and 2% of the encoded byte stream, between 0.3% and 1.8%, between 0.5% and 1.6%, between 0.7% and 1.4%, and between 0.9% and 1.2%. [00303] In some embodiment, the first sequence scan includes at least 0.1% of the encoded byte stream, at least 0.3% of the encoded byte stream, at least 0.5% of the encoded byte stream, at least 0.7% of the encoded byte stream, at least 0.9% of the encoded byte stream, at least 1.1% of the encoded byte stream, at least 1.3% of the encoded byte stream, at least 1.5% of the encoded byte stream, at least 1.7% of the encoded byte stream, or at least 1.9% of the encoded byte stream. [00304] In some embodiment, the first sequence scan includes at most 0.1% of the encoded byte stream, at most 0.3% of the encoded byte stream, at most 0.5% of the encoded byte DB1/ 147040794.2 77 Attorney Ref. No.: 115834-5037-WO stream, at most 0.7% of the encoded byte stream, at most 0.9% of the encoded byte stream, at most 1.1% of the encoded byte stream, at most 1.3% of the encoded byte stream, at most 1.5% of the encoded byte stream, at most 1.7% of the encoded byte stream, or at most 1.9% of the encoded byte stream. [00305] Block 462. Referring to block 462, in some embodiments, the first sequence scan includes less than 5% of the final terminal scan. In some embodiments, the first sequence scan includes between 0.5% and 4% of the final terminal scan, between 1% and 2.5%, between 1.5% and 1.6%, between 0.7% and 1.4%, and between 0.9% and 1.2%. [00306] In some embodiment, the first sequence scan includes at least 0.1% of the final terminal scan, at least 0.3% of the final terminal scan, at least 0.5% of the final terminal scan, at least 0.7% of the final terminal scan, at least 0.9% of the final terminal scan, at least 1.1% of the final terminal scan, at least 1.3% of the final terminal scan, at least 1.5% of the final terminal scan, at least 1.7% of the final terminal scan, or at least 1.9% of the final terminal scan. [00307] In some embodiment, the first sequence scan includes at most 0.1% of the final terminal scan, at most 0.3% of the final terminal scan, at most 0.5% of the final terminal scan, at most 0.7% of the final terminal scan, at most 0.9% of the final terminal scan, at most 1.1% of the final terminal scan, at most 1.3% of the final terminal scan, at most 1.5% of the final terminal scan, at most 1.7% of the final terminal scan, or at most 1.9% of the final terminal scan. [00308] Block 464. Referring to block 464, in some embodiments, the encoded byte stream 114 is associated with a first file size of the final terminal scan in the plurality of sequence scans. Moreover, a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan. [00309] Block 466. Referring to block 466, in some embodiments, the method 400 includes terminating the communicating of the encoded byte stream 114. For instance, in some embodiments, the method 400 terminates communicating the encoded byte stream 114 via the communication network 186 in accordance with a determination that decoding of the encoded byte stream 114 at the second computer system satisfies a first sequence scan in the plurality of sequence scans. In some embodiments, the method 400 terminates communicating the encoded byte stream 114 via the communication network 186 in accordance with a determination that decoding of the encoded byte stream 114 at the second DB1/ 147040794.2 78 Attorney Ref. No.: 115834-5037-WO computer system the first sequence scan matches or exceeds the first training resolution. In some embodiments, this determination is made at the first computer system. In some embodiments, this determination is made at the second computer system. In some embodiments, this determination is made prior to initiation of the communicating of the encoded byte stream. However, the present disclosure is not limited thereto. [00310] Accordingly, in some embodiments, the method 400 obtains a training data set including the first plurality of graphical data 108-1 in order to encode an encoded byte stream 114 of second one or more graphical data 108-2 in accordance with a request for an evaluation of the second one or more graphical data. From this, the method 400 allows for an optimal subset of the encoded byte stream 114, which is associated with the first training resolution, to be communicated via the communication network 186 without negatively impacting performance of a respective model 118 to which the second one or more graphical data 108-2 is applied. In some embodiments, the first sequence scan is identified in accordance with a requirement to satisfy a desired threshold performance so there is no statistically significant difference in the model performance between the final terminal scan and the first sequence scan in the plurality of sequence scans. [00311] Now that a general method 400 for optimizing decoding of graphical data has been described in accordance with various embodiments of the present disclosures, details regarding some processes in accordance with Figures 5A through 5E will be described. [00312] Various modules in a memory 192 of a graphical data system 200 (e.g., graphical data store 106 of Figure 2A, model library 116 of Figure 2B, codec module 122 of Figure 2B, etc.) and/or a memory 292 of a client device 300200 (e.g., model library 216 of Figure 3, codec module 222 of Figure 3, etc.) perform certain processes of the methods of the present disclosure, unless expressly stated otherwise. Furthermore, it will be appreciated that the processes of a method of the present disclosure can be encoded in a single module or any combination of modules. [00313] Referring now to Figures 5A through 5E, there is depicted a flowchart illustrating an exemplary method 500 in accordance with some embodiments of the present disclosure. In the flowchart, the preferred parts of the method are shown in solid line boxes, whereas additional, optional, or alterative parts of the method are shown in dashed line boxes. [00314] In some embodiments, the method 500 optimizes encoding of graphical data (e.g., graphical data 108 of Figure 2A), communication of graphical data (e.g., encoded byte stream DB1/ 147040794.2 79 Attorney Ref. No.: 115834-5037-WO 114 of Figure 2A), decoding of graphical data, or a combination thereof that is communicated via a communication network (e.g., communication network 186 of Figure 1). In particular, in some embodiments, the method 500 communicates an encoded byte stream (e.g., second encoded byte stream 114-2 of Figure 2A) in accordance with a prior training using a different graphical data set (e.g., first graphical data 108-1 of Figure 2A), a use case defined by a request for the encoded byte stream 114 (e.g., defined by a request received from a second computer system such as client device 300 of Figure 3), an aspect of a client device 300 associated with the request for the encoded byte stream 114, a required resolution for using the graphical data 108 of the encoded byte stream 114, an aspect of a model (e.g., model 118 of Figure 2B) to which the graphical data 108 is applied, or a combination thereof. From this, in some embodiments, the method 500 allows for communicating only an optimal resolution, or corresponding scan in a plurality of sequence scans of the encoded byte stream 114 that matches a desired threshold performance. Accordingly, in some embodiments, the method 500 allows or optimizing cost (e.g., data storage costs, data capacity costs, etc.), network bandwidth, turnaround time, or a combination thereof for communicating graphical information through the encoded byte stream 114. Moreover, in some embodiments, the method 500 allows for terminating communication of the encoded byte stream in accordance with a determination that a first resolution of the encoded byte stream decoded at the second computer system satisfies a threshold resolution, such as by matching or exceeding a first resolution. However, the present disclosure is not limited thereto. [00315] In various embodiments, the method 500 is provided at a computer system (e.g., graphical data system 200 of Figures 2A and 2B, client device 300 of Figure 3, etc.). The computer system 200 includes one or more processors (e.g., CPU 172 of Figure 2A, CPU 272 of Figure 3, etc.) and a memory (e.g., memory 192 of Figures 2A and 2B, memory 392 of Figure 3, etc.), such a first memory coupled to the one or more processors. In some embodiments, the memory includes one or more programs configured to be executed by the one or more processors (e.g., graphical data store 106 of Figure 2A, model library 116 of Figure 2B, codec module 122 of Figure 2B, model library 216 of Figure 3, codec module 222 of Figure 3, etc.). Accordingly, in some such embodiments, the method 500 requires utilization of a computer system, such as in order to evaluate the graphical data 108 and communicate the graphical data 108 via the communication network 186, and, therefore, cannot be mentally performed. DB1/ 147040794.2 80 Attorney Ref. No.: 115834-5037-WO [00316] Furthermore, in some embodiments, the method 500 allows for parallelism, which allows for encoding (e.g., using first encoder 126-1 of Figure 2B), communicating an encoded byte stream, decoding the encoded byte stream (e.g., using second decoder 126-2 of Figure 2B), or a combination thereof simultaneously using a plurality of graphical data 108 and/or using a plurality of CPUs 172. However, the present disclosure is not limited thereto. [00317] Block 502. Referring to block 502, in some embodiments, the method 500 includes obtaining first graphical data 108-1 (e.g., a first image, a set of images, a 3D model, etc.). For instance, in some embodiments, the first graphical data 108-1 is obtained in electronic form via the communication network 186 and/or retrieved for the memory 192 of the computer system 200. In some embodiments, the first graphical data 108-1 is captured prior to being obtained by the computer system 200. However, the present disclosure is not limited thereto. [00318] In some embodiments, the first one or more graphical data 108-1 is defined by a first plurality of characteristics (e.g., first plurality of graphical data 110-1 of Figure 2A, first plurality of characteristics 110 of method 500 of Figures 5A-5E, etc.). In some embodiments, the first plurality of characteristics 110 includes a sampling resolution of the first graphical data 108, such as a minimum resolution associated with an image in the first graphical data 108, a mode resolution associated with the first graphical data 108, or the like. Moreover, in some embodiments, the first plurality of characteristics 110 further includes a first modality associated with a capture of the first graphical data 108-1, such as a first X-Ray modality or a second thermal imaging modality. [00319] Block 504. Referring to block 504, in some embodiments, the first graphical data 108-1 includes one or more images taken of a first biological sample. In some embodiments, the first graphical data is as exemplified by at least block 406. [00320] Block 506. Referring to block 506, in some embodiments, the first one or more graphical data 108-1 includes one or more images taken of a second biological sample different than the first biological sample. In some embodiments, the first graphical data is as exemplified by at least block 408. [00321] Block 508. Referring to block 508, in some embodiments, the sampling resolution of the first one or more graphical data 108-1 includes the minimum resolution associated with the first one or more graphical data 108-1. In some embodiments, the first graphical data is as exemplified by at least block 414. DB1/ 147040794.2 81 Attorney Ref. No.: 115834-5037-WO [00322] Block 510. Referring to block 510, in some embodiments, the first modality (e.g., third characteristic 112-3 of Figure 2A) includes a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof. In some embodiments, the first modality is as exemplified by at least block 416. [00323] Block 512. Referring to block 512, in some embodiments, the first plurality of characteristics 110 further includes one or more ground truth labels (e.g., fourth characteristic 112-4 of Figure 2A, etc.) with respect to the first modality for the first one or more of graphical data 108-1. In some embodiments, the first plurality of characteristics is as exemplified by at least block 424. [00324] Block 514. Referring to block 514, in some embodiments, the first plurality of characteristics 110 further includes a respective capacity of the communication network 186, such as a first bandwidth for receiving data through the communication network 186 and/or a second bandwidth for communicating data through the communication network 186. In some embodiments, the first plurality of characteristics is as exemplified by at least block 436. [00325] Block 516. Referring to block 516, in some embodiments, the first plurality of characteristics 110 further includes a respective capacity of the second computer system 300. In some embodiments, the first plurality of characteristics is as exemplified by at least block 438. [00326] Block 518. Referring to block 518, in some embodiments, the method 500 further includes forming a byte stream map (e.g., byte stream map 600 of Figure 6) for the first one or more graphical data 108-1. In some embodiments, the byte stream map 600 includes a plurality of sequence scans (e.g., sequence scans 1010 of method 400 of Figures 4A-4E, scans 1010 of Figure 7, scans 1010 of Figure 8, scans 1010 of Figure 9, scans 1010 of Figure 10, scans 1010 of Figure 11, scans 1010 of Tables 1-14, etc.) through the sampling resolution, which allows an end-user to view, at a highest resolution and/or optimal resolution, the sampling resolution of the first graphical data 108-1 when communicated to a remote device, such as the second computer system 300. In some embodiments, each respective sequence scan 1010 in the plurality of sequence scans is associated with a unique rank in a rank order. In some embodiments, each rank order defines a sequence of respective sequence scans 1010 in the plurality of sequence scans. For instance, in some embodiments, the rank order defines DB1/ 147040794.2 82 Attorney Ref. No.: 115834-5037-WO a structured organized set that assigns each unique rank in the rank order to a position in the rank order. In some embodiments, the rank order is based on a value of a corresponding resolution. Furthermore, in some embodiments, each sequence scan 1010 in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans. For instance, in some embodiments, each subsequent sequence scan 1010 in the plurality of sequence scans 1010 has an increasing level of quality (e.g., first characteristic 112-1) and resolution (e.g., second characteristic 112-2), such that the complete reconstruction through the final terminal scan 1010 is lossless. In some embodiments, the plurality of sequence scans include N scans, in which N is a positive integer, which allows for reconstruction of ^^^^ + 1 scans 1010, where the resolution of each scan 1010 is a factor of 2 ^^^^+1 of the original image 108 resolution. [00327] For instance, in some embodiments, each graphical data 108 in the first graphical data is associated with a unique entry or data elements of the byte stream map 600. For instance, in some embodiments, each sequence scan 1010 is mapped via the byte stream map 6000 to the partial byte stream 114 necessary to decode a corresponding graphical data 108 at the corresponding resolution, or decomposition level, of the sequence scan 1010. In some embodiments, an entry, or node, in the byte stream map 600 is created by generating a plurality of tile-part markers in the encoded byte stream 114. In some embodiments, each entry, or scan 1010, is associated with at least one marker. In some embodiments, the tile- park marker allow for graphical data 108 to be divided into a plurality of tiles, or tessellations, in which each tile is encoded in the byte stream map 600, as opposed to encoded the entirety of the graphical data 108. In some embodiments, each tile-part marker is formed in a rectangular shape. In some embodiments, each tile park marker is tessellated. As used herein, the term “tessellated” means a tiling of a plane using one or more geometric shapes, hereinafter “tiles,” with no overlaps or gaps therebetween. For instance, in some implementations, the tessellation of a region refers to the tiling of a plane, or surface, that defines the region (e.g., the tessellation of a plane in a 2D polar coordinate system, the tessellation of a surface in a 3D spherical coordinate system). For instance, in some embodiments, an initial terminal scan 1010 is associated with a first marker and a final terminal scan 1010 is associated with a second marker. As a non-limiting example, in some embodiments, the plurality of tile-part markers include a first tile-part marker identified by DB1/ 147040794.2 83 Attorney Ref. No.: 115834-5037-WO byte 0xFF90 and a second tile-part marker identified by byte 0xFFD9 of the encoded byte stream 114, which allows for indicating data in the encoded byte stream for a particular tile- part marker. In some embodiments, a number of sequence scans in the plurality of sequence scans is determined as a factor, or function, of 2 ^^^^+1 of the graphical data’s 108 full, or original resolution. In some embodiments, the inclusion of tile-part markers in the encoded byte stream map 600 enables graphical data 108 to be communicated at a specific scan 1010 level in the plurality of sequence scans 1010 using a portion of the encoded byte stream 114 without requiring communicating the complete byte stream 114. Accordingly, in some embodiments, the byte stream map 600 allows for navigating through the plurality of sequence scans 1010 to access an optimal scan (e.g., block 456 of method 400, block 458 of method 500, etc.) when communicating and/or decoding the encoded byte stream 114 associated with the byte stream map 600. Moreover, in some embodiments, the byte stream map 600 reduces an amount of computational processing required by the first computer system 200 and/or the second computer system 300 by not requiring communication and/or decoding of other scans 1010 in the plurality of sequence scans 1010, such as those that exceed the desired threshold performance. However, the present disclosure is not limited thereto. [00328] In some embodiments, the byte stream map 600 allowed for matching tile-part divisions by resolution (e.g., forming the byte stream map 600) in accordance with the tile- part marker for an encoded byte stream 114 of graphical data 108. For instance, in some embodiments, provided that the encoded byte stream 114 to reconstruct the image at a sub- resolution is available, the graphical data is decoded at the scan level and all preceding scan levels even in the absence of the complete encoded byte stream 114 based on the byte stream map 600. This enables quick access to smaller resolution versions of the graphical data (e.g., for thumbnails) based on the byte stream map 600 and the scans 1010 therein, thus minimizing data duplication. [00329] Block 520. Referring to block 520, in some embodiments, the first one or more graphical data 108 includes 3D graphical data and each scan in the plurality of sequence scans of the 3D graphical data includes a respective 2D layer of the 3D graphical data. In some embodiments, the first graphical data is as exemplified by at least block 438. [00330] Block 522. Referring to block 522, in some embodiments, the first one or more graphical data 108 includes one or more digital images, one or more digital videos, one or DB1/ 147040794.2 84 Attorney Ref. No.: 115834-5037-WO more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. [00331] Block 524. Referring to block 524, in some embodiments, the plurality of sequence scans 1010 includes between 2 scans and 100 scans. In some embodiments, the plurality of sequence scans is as exemplified by at least block 448. [00332] Block 526. Referring to block 526, in some embodiments, the plurality of sequence scans 1010 includes N scans, in which ^^^^ is� ^^^^ ^^^^ ^^^^ ^^^^ 2� ^^^^ ^^^^ ^^^^ ^^^^ ^^^^√2��. In some embodiments, M is a number of digital assets associated with the second one or more graphical data. In some embodiments, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data. Moreover, in some embodiments, ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. [00333] In some embodiments, the plurality of sequence scans 1010 includes N scans, in which ^^^^ = ^^^^ ^^^^ 2 − 1, M is a number of digital assets associated with the second one
Figure imgf000087_0001
or more graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. [00334] In some embodiments, the plurality of sequence scans is as exemplified by at least block 450. [00335] In some embodiments, the plurality of sequence scans 1010 includes N scans, in which ^^^^ is determined for each graphical data in the one or more graphical data 108 separately prior to forming the encoded byte stream map 600 and/or encoded byte stream 114, which ensured that the corresponding resolution of a sequence scan 1010 remained consistent across medical images with different resolutions as follows ^^^^ =�log ^^^^ ^^^^ ^^^^ ( ^^^^, ^^^^) 2 ^^^^ � where ^^^^ and ^^^^ are a horizontal resolution and a vertical resolution, respectively, of the graphical data 108, and ^^^^ is a lower bound for the resolution of the initial terminal scan. [00336] Block 528. Referring to block 528, in some embodiments, In some embodiments, the plurality of sequence scans includes N scans, in which ^^^^ = ^^^^ ^^^^ ^^^^ − 1, M is a
Figure imgf000087_0002
number of digital assets associated with the second one or more graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and DB1/ 147040794.2 85 Attorney Ref. No.: 115834-5037-WO ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. [00337] In some embodiments, the plurality of sequence scans is as exemplified by at least block 452. [00338] Block 530. Referring to block 530, in some embodiments, the first plurality of characteristics 110 includes a sampling resolution of the first one or more graphical data 108. In some embodiments, the plurality of sequence scans 1010 includes N scans based, at least in part, on the sampling resolution of the first one or more graphical data 108. In some embodiments, the first plurality of characteristics is as exemplified by at least block 454. [00339] Block 532. Referring to block 532, in some embodiments, the method 500 includes receiving, from a second computer system 300, a request to evaluate the first one or more graphical data 108 at a first resolution. In some embodiments, the first resolution is selected in accordance with a desired threshold performance from performance data. In some embodiments, the performance data includes a plurality of training resolutions. In some embodiments, each respective training resolution in the plurality of training resolutions is less than the sampling resolution. Moreover, in some such embodiments, each respective training resolution is associated with a corresponding threshold performance for a trained feature extraction model 118 in a plurality of threshold performances. In some embodiments, each threshold performance in the plurality of threshold performances associated with the first modality associated with a capture of a second plurality of graphical data 108. [00340] Block 534. Referring to block 534, in some embodiments, the second plurality of graphical data 108 includes two-dimensional (2D) graphical data, three-dimensional (3D) graphical data, four-dimensional (4D) data, or a combination thereof. In some embodiments, the second graphical data is as exemplified by at least block 408. [00341] Block 536. Referring to block 536, in some embodiments, the second plurality of graphical data 108 includes one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. In some embodiments, the second graphical data is as exemplified by at least block 410. [00342] Block 538. Referring to block 538, in some embodiments, the second plurality of graphical data 108 includes 50 or more digital images, 100 or more digital images, 1,000 or DB1/ 147040794.2 86 Attorney Ref. No.: 115834-5037-WO more digital images, 10,000 or more digital images, or 100,000 or more digital images. In some embodiments, the second graphical data is as exemplified by at least block 412. [00343] Block 540. Referring to block 540, in some embodiments, prior to the receiving C), the method 500 further includes training the feature extraction model 118 against the one or more ground truth labels. In some embodiments, the ground truth label and/or training is as exemplified by at least blocks 424-426. [00344] Block 542. Referring to block 542, in some embodiments, the desired threshold performance includes a threshold inter-image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation. In some embodiments, the desired threshold performance is as exemplified by at least block 440. [00345] Block 544. Referring to block 544, in some embodiments, the feature extraction model 118 includes a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof. In some embodiments, the feature extraction model is as exemplified by at least block 420. [00346] Block 546. Referring to block 546, in some embodiments, the feature extraction model 118 includes a neutral network model, a support machine mode, a Naïve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph-based model, or a combination thereof. In some embodiments, the feature extraction model is as exemplified by at least block 422. [00347] Block 548. Referring to block 548, in some embodiments, the method 500 further includes matching the first resolution to a first sequence scan 1010-1 in the plurality of sequence scans 1010 in accordance with the byte stream map 600. In some embodiments, the matching is as exemplified by at least block 456. [00348] Block 550. Referring to block 550, in some embodiments, the first codec 124-1 includes a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof. In some embodiments, the first codec 124-1 is as exemplified by at least block 444. [00349] Block 552. Referring to block 552, in some embodiments, the first codec 124-1 is selected by the second computer system 300. In some embodiments, the first codec 124-1 is as exemplified by at least block 446. DB1/ 147040794.2 87 Attorney Ref. No.: 115834-5037-WO [00350] Block 554. Referring to block 554, in some embodiments, the first sequence scan 1010-1 includes less than 5% of the encoded byte stream 114. In some embodiments, the first sequence scan is as exemplified by at least block 460. [00351] Block 556. Referring to block 556, in some embodiments, the first sequence scan 1010-1 includes less than 5% of a final terminal scan. In some embodiments, the first sequence scan is as exemplified by at least block 462. [00352] Block 558. Referring to block 558, in some embodiments, the method 500 includes encoding, in accordance with a corresponding encoder 126 of a first codec 124-1, the first one or more graphical data 108, thereby forming an encoded byte stream 114. In some embodiments, the encoded byte stream 114 includes the plurality of sequence scans through at least the first resolution. In some embodiments, the encoded byte stream 114 is as exemplified by at least block 442. [00353] Block 560. Referring to block 560, in some embodiments, the encoded byte stream 114 is associated with a first file size of the final terminal scan 1010 in the plurality of sequence scans, and a second file size of the first sequence scan 1010 in the plurality of sequence scans is less than the first file size of the final terminal scan 1010. In some embodiments, the encoded byte stream 114 is as exemplified by at least block 464. [00354] Block 562. Referring to block 562, in some embodiments, the method 500 further includes communicating, via a communication network 1016, in accordance with the first codec 124-1, the encoded byte stream 114 and one or more instructions to terminate decoding of the encoded byte stream 114 at the second computer system 300 to the second computer system 300. In some embodiments, the one or more instructions to terminal decoding execute when the decoding decodes the first sequence scan 1010 in the plurality of sequence scans at the second computer system 300, thereby optimizing decoding of graphical data. In some embodiments, the communicating is as exemplified by at least block 458. [00355] Example 1 [00356] Experimental Setup [00357] This Example evaluated the systems and methods of the present invention (e.g., graphical data system 200 of Figures 2A and 2B, method 400 of Figures 4A through 4E, etc.) in two experimental setups. The first experiment employed a 2D classification model (e.g., first model 118-1 of Figure 2B) that detected abnormalities in a second one or more graphical data (e.g., graphical data 108-2 of Figure 2A) associated with a first modality (e.g., first DB1/ 147040794.2 88 Attorney Ref. No.: 115834-5037-WO characteristic 112-1) of the thorax using chest x-ray (CXR) graphical data. The second experiment employed a 3D segmentation model (e.g., second model 118-2 of Figure 2B) that segmented a third plurality of graphical data (e.g., graphical data 108-3 of Figure 2A) associated with a second modality (e.g., second characteristic 112-2) of the liver and spleen using computed tomography (CT) data. In both experiments, the models 118 were trained using a first plurality of graphical data (e.g., graphical data 108-1 of Figure 2A) that was local to a first computer system (e.g., graphical data system 200 of Figures 2A and 2B), validated using held-out internal test sets, and evaluated on unseen (e.g., discrete from the first plurality of graphical data) graphical data received from (e.g., via communication network 186) a hospital system (e.g., client device 300 of Figure 3). Furthermore, in both experiments, the graphical data 106 communicated, decode time using a decoder 128, and throughput (e.g., a number of images processed by a respective model 118 per unit time (e.g., second)), which were measured as data metrics and compared to an original non-high throughput JPEG 2000 (HTJ2K) version of the respective plurality of graphical data 108. [00358] Classification [00359] A trained multi-label DenseNet121 model 118 architecture was utilized, initialized with ImageNet weights, on the NIH Chest X-Ray (“NIH”) dataset 108 (e.g., first graphical data of block 404 of method 400, second graphical data of block 532 of method 500, etc.) using transfer learning. See Huang et al., 2017, “Densely Connected Convolutional Networks,” Proceedings of the IEEE conference on computer vision and pattern recognition. pp.4700-4708; Wang et al., 2017, “Chestx-ray8: Hospital-scale Chest X-ray Database and Benchmark son Weakly-supervised Classification and Localization of Common Thorax Diseases,” Proceedings of the IEEE conference on computer vision and pattern recognition. pp.2097–2106, each of which is hereby incorporated by reference in its entirety for all purposes. The NIH graphical data 108 included 14 disease labels (e.g., ground truth label characteristics 112) with 112,120 frontal-view CXRs (e.g.g., chest x-ray modality characteristic 112) from 30,805 patients (e.g., first graphical data 108-1 of Figure 2A, first graphical data 108 of block 404 of method 400 of Figures 4A-4E, etc). The graphical data 108 was randomly divided into training (70%), validation (10%), and testing (20%, N = 22,330) splits and ensured no patient appeared in more than one split. All graphical data 108 was downsampled to a resolution of 224x224 and normalized between 0 and 1. Graphical data 108 within the held-out NIH test set were originally encoded as a PNG filetype (e.g., first PNG code 124-1 using first encoder 126-1, first codec 124-1 of block 442 of method 400 DB1/ 147040794.2 89 Attorney Ref. No.: 115834-5037-WO of Figure 4A-4E, first code 124-1 of block 558 of method 500 of Figures 5A-5E, etc.) with a normalized resolution of 1024x1024. The NIH test set 108 was progressively encoded 128 using lossless HTJ2K codec (e.g., second codec 124-1 of Figure 2B), which included 5 scans 1010 in a plurality of sequence scans that yielded a maximum compression ratio of 256:1 for the corresponding byte stream map 600 (e.g., byte stream map 60 of method 500 of Figures 5A-5E). A first resolution associated with an optimal scan 1010 in the plurality of sequence scans was identified (e.g., matching of block 456 of method 500, matching of block 548 of method 500, etc.) and evaluated the model 118 across each scan 1010 of HTJ2K encoded byte stream 114 and original PNG graphical data 108. [00360] The optimal task-specific byte stream scan 1010 obtained using the NIH data 108 was evaluated on a second one or more graphical data (e.g., an external dataset obtained by randomly sampling N = 10,000 CXRs from the MIMIC-CXR-JPG dataset, first graphical data of block 404 of method 400, second graphical data of block 532 of method 500, etc.). See Goldberger et al., 2000, “"PhysioBank, PhysioToolkit, and PhysioNet: Components of a New Research Resource for Complex Physiologic Signals,” Circulation, 101(23), pg. e215- e220; Johnson, et al., 2019, “MIMIC-CXR-JPG, A Large Publicly Available Database of Labeled Chest Radioraphs,” arXiv preprint, arXiv 1901.07042, print, each of which is hereby incorporated by reference in its entirety for all purposes. All uncertain ground truth labels 112 were treated as negatives. Unlike the NIH test set, graphical data 108 within the MIMIC test set were originally encoded using the original, and heavily optimized, JPEG standard 124 with resolutions of scans 1010 ranging from 1470x1461 to 4280x3520 (e.g., first encoder 126-1 of Figure 2B). The MIMIC test set was progressively encoded 126 using lossless HTJ2K code 124, which included 7 scans 1010 in a plurality of sequence scans, to yield a maximum compression ratio of 4096:1. Performance of a respective model 118 was evaluated across each scan 1010 of the HTJ2K encoded byte stream 114 and original JPEG data 108 by a desired threshold performance using a binary cross entropy loss (BCE) (e.g., a first threshold performance criterion), an area under the receiver operating characteristic 112 (AUROC) (e.g., a second desired threshold performance), and an area under the precision recall curve (AUPRC) (e.g., a third threshold performance criterion). In a multi-ground truth label scenario, the AUROC and AUPRC scores 112 were defined as the mean score across all disease labels. A 95% confidence interval (CI) for the AUROC scores was determined and compared using bootstrapping and a paired two-tailed t-test (e.g., third model 118-3 of Figure 2B). Statistical significance was defined as p < 0.05. DB1/ 147040794.2 90 Attorney Ref. No.: 115834-5037-WO [00361] Segmentation: [00362] A trained 3D-UNet model 118 used the MONAI framework for segmentation of liver and spleen graphical data 108 and used the corresponding graphical datasets 108 from the Medical Segmentation Decathlon (MSD) training sets (e.g., first graphical data of block 404 of method 400, second graphical data of block 532 of method 500, training data of Example 3, etc.). See Cardoso et al., 2022, “Monai: An open-source framework for deep learning in healthcare,” arXiv preprint arXiv:2211.02701, print; Çiçek et al., 2016, “3D U- Net: Learning Dense Volumetric Segmentation from Sparse Annotation,” Medical Image Computing and Computer-Assisted Intervention–MICCAI 2016: 19th International Conference, Athens, Greece, Proceedings, Part II(19), pp.424-432, each of which is hereby incorporated by reference in its entirety for all purposes. In this Example, liver tumor annotations were discarded. Volumes for the held-out MSD liver test set (N = 25) and spleen test set (N = 9) were sampled prior to training of a respective model 118. The model 118 architecture included 5 two-convolutional residual units with down/up-sampling by a factor of 2 at each unit. All graphical data 108 volumes were resized to a resolution of 256x256x128 and voxel values normalized between 0 and 1 Random. Foreground patches of resolution 128x128x32 were extracted from each volume such that a center voxel of each patch belonged to either a foreground class or a background class. [00363] Graphical data 108 within the held-out MSD liver and spleen test sets were originally encoded as NifTI 124 volumes with a normalized resolution of 512x512xN. One limitation of the HTJ2K standard 124 is its inability to encode 3D graphical data 108. As a workaround, for all volumes, 3D graphical data was progressively encoded by each 2D slice using lossless HTJ2K 124, which yielded four scans 1010 in a plurality of sequence scans, which yielded a maximum compression ratio of 64:1. All slices were decoded (e.g., using a first decoder 128-1 of Figure 2B) individually, reconstructed into a volume, and evaluated with the model 118, along with the original NifTI volumes. The optimal byte stream scan 110 was evaluated and obtained from internal validation on an external dataset including N = 30 volumes from the Beyond the Cranial Vault (BTCV) dataset. See Landman et al., 2015, “Miccai Multiatlas Labeling Beyond the Cranial Vault–Workshop and Challenge,” MICCAI Multi-Atlas Labeling Beyond Cranial Vault – Workshop Challenge, 5, pg.12, which is hereby incorporated by reference in its entirety for all purposes. All volumes were similarly encoded as NifTI 124 with a normalized resolution of 512x512xN, and processed and evaluated similarly. The mean DICE score between the ground truth and model 118 DB1/ 147040794.2 91 Attorney Ref. No.: 115834-5037-WO segmentation was measured as the model 118 metric and compared using a paired t-test. Statistical significance was again defined as p < 0.05. [00364] Results [00365] On the held-out NIH test set, model 118 performance for all progressive scans was comparable to using full resolution PNG graphical data 108, and only utilized a fraction of the amount of data 108 communicated and decoding time (apart from scan 1) (Figure 7). For instance, scan 2 yielded a comparable mean AUROC of 0.81 (p = 0.74), communicated 97.25% less data 108, took 95.60% less time to decode, and yielded a factor of 9.90 increase in throughput. Despite PNG coding 124 being heavily optimized due to its wide adoption, decoding the full resolution images from the complete byte stream required 4.41% less data to be communicated and took 7.40% less time to decode using decoder 128. [00366] On the MIMIC external test set, apart from the first two scans in the plurality of sequence scans, similar, if not better, performance was observed for a respective model 118 for all progressive scans 1010 (Figure 8). For scan 3 in the plurality of sequence scans, the a mean AUROC of 0.75 (p = 0.77) was determined, which communicated 99.01% less data, took 98.58% less time to decode (e.g., using the first decoder 128-1 of Figure 2B), and yielded a 27.43 factor increase in throughput. Due to the efficient and heavily optimized nature of the original JPEG standard, scan 7 in the plurality of sequence scans required a factor of 1.78 more data to be communicated and took a factor of 1.16 more time to decode, which impacted the overall throughput. [00367] On the held-out MSD liver and spleen test sets, compression significantly impacted segmentation tasks. Despite achieving a comparable Dice similarity coefficient (DICE score) of 0.94 ± 0.02 from the scan 3 in the plurality of sequence scans and the scan 4 in the plurality of sequence scans for liver segmentation, spleen segmentation performed poorly for all progressive scans in the plurality of sequence scans apart from the final full resolution scan (e.g., final terminal scan in the plurality of sequence scans). However, across both tasks, the superiority of HTJ2K over NifTI compression was demonstrated. On the MSD liver test set, lossless HTJ2K 124 data required 76.94% less data to be communicated, 76.33% less time to decode, and a yielded factor of 1.89 increase in throughput. Similarly, the HTJ2K 124 encoded MSD spleen test set required 67.66% less data to communicate, 86.56% less time to decode, and yielded a factor of 1.66 increase in throughput (Figure 9). DB1/ 147040794.2 92 Attorney Ref. No.: 115834-5037-WO [00368] On the BTCV external test set, similar results as the MSD test sets were observed, in that liver segmentation performed well for volumes with higher compression ratios while spleen segmentation was severely impacted by it (Figure 10). However, the scan 3 in the plurality of sequence scans yielded excellent DICE scores of 0.88±0.09 and 0.81±0.17 on liver and spleen, respectively, which required 90.32% less data to communicate, 90.26% less time to decode, and yielded a factor of 1.61 increase in throughput. Furthermore, lossless HTJ2K volumes of graphical data 108 required 73.12% less data to communicate and 81.20% less time to decode when compared to the original NifTI 124 compressed volumes. [00369] Utilizing a respective classification model 118 observed reduced the data communicated by 99.01% and decoding time by 98.58%, and increased throughput by a factor of 27.43. For segmentation, data communicated was reduced by 90.32%, decoding time by 90.26%, and increased throughput by a factor 1.61. [00370] Accordingly, the ability to significantly reduce storage requirements for graphical data, bandwidth requirements for communicating the graphical data, and time to decode the graphical data was demonstrated, which increased overall throughput when utilizing the graphical data. Due to the robustness of deep learning models 118 for CXR classification, high compression ratios were achieved with a non-significant impact to performance for a respective model 118. See Koff et al., 2006, An Overview of Digital Compression of Medical Images: Can We Use Lossy Image Compression in Radiology?,” Journal-Canadian Association of Radiologists, 57(4), pg.211; Sabottke et al., 2020, The Effect of Image Resolution on Deep Learning in Radiography,” Radiology: Artificial Intelligence, 2(1), pg. e190015; Saffor et al, 2001, “A Comparative Study of Image Compression Between JPEG and Wavelet,” Malaysian Journal of Computer Science, 14(1), pg.39-45, each of which is hereby incorporated by reference in its entirety for all purposes. Despite excellent performance on liver segmentation models 118, correct segmentation of the spleen proved difficult. Prior literature had also suggested segmentation tasks are more sensitive to compression. See Liu et al., 2019, “Machine Vision Guided 3D Medical Image Compression for Efficient Transmission and Accurate Segmentation in the Clouds,” Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pg.12687-12696, which is hereby incorporated by reference in its entirety for all purposes. Apart from being a much harder task due to a corresponding small size, a lack of using random intensity shift in the data augmentation step, resulted in a poor generalizability. See Zhang et al., 2020, “Generalizing Deep Learning for Medical Image Segmentation to Unseen Domains via Deep DB1/ 147040794.2 93 Attorney Ref. No.: 115834-5037-WO Stacked Transformation,” IEEE Transactions on Medical Imaging, 39(7), pg.2531-2540, which is hereby incorporated by reference in its entirety for all purposes. Graphical data volumes within the intensity range of the original training data correctly segmented the spleen. Despite these results, the superiority of HTJ2K 124 for communicating image volumes by significantly reducing communication and decoding overheads was demonstrated, while increasing throughput. Furthermore, the code 124 resulted in faster turnaround times, reduced overall cost of data storage and communication, and failed to negatively impact clinical decision making. [00371] Example 2 [00372] In this Example, encoding parameters (e.g., a number of scans in a plurality of sequence scans) were determined by analyzing a plurality of graphical data (e.g., first graphical data 108-1 of Figure 2A, second graphical data 108-2 of Figure 2A, graphical data of Example 1, graphical data of Example 3, etc.). A number of scans 1010, N, in the plurality of sequence scans was determined by evaluating the entirety of a respective plurality of graphical data (e.g., the entire second one or more graphical data 108-2), such that ^^^^ was�log2� ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^√2��. ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} was a native resolution of the graphical data, ^^^^ was the number of graphical data within the set, and ^^^^ ^^^^ ^^^^ ^^^^ was the minimum resolution of the first scan. [00373] Accordingly, the optimal set of scans 1010 to encode for each graphical data within the second one or more graphical data was determined, which ensured the resolution of the first scan (e.g., an initial terminal scan in the plurality of sequence scans) satisfied a threshold ^^^^ ^^^^ ^^^^ ^^^^, such as the initial terminal scan exceeded ^^^^ ^^^^ ^^^^ ^^^^. The encoded byte stream encoded by the code 124 ensured normalization of each resolution associated with a corresponding scan in the plurality of sequence scans, which included the initial terminal sequence scan and any subsequent scan to prevent resolution disparities. For graphical data with smaller resolution, the graphical data was encoded 126 using fewer scans 1010, while graphical data 108 with higher resolutions using more scans. [00374] For the MIMIC graphical data, a first image with resolution 1470x1461 in the second one or more graphic data was encoded with 6 scans in a first plurality of sequence scans with an initial terminal scan with resolution of 46x46. A second image with resolution 4280x3520 in the second one or more graphic data was encoded with 7 scans in a second plurality of DB1/ 147040794.2 94 Attorney Ref. No.: 115834-5037-WO sequence scans with an initial terminal scan with resolution of 67x55. This approach ensured that all images within the second one or more graphical data yielded an initial terminal scan with a resolution within a range of a corresponding resolution of the initial terminal scan of the first and second images, such as range of [~45, ~90]. The corresponding resolution of all subsequent scans in the plurality of sequence scans was normalized with a fixed range of resolutions. [00375] Example 3 – Development of a Framework for Medical Image Compression [00376] Progressive Encoding [00377] A progressive encoding, decoding, and communication framework (e.g., as illustrated in Figures 13A-B) was based on lossless progressive encoding, a type of compression offered by the JPEG 2000 standard (e.g., first codec 124-1 of Figure 2B, first code 124-1 of Figure 3, first codec 124-1 of block 422 of method 400 of Figures 4A through 4E, first codec 124-1 of block 558 of method 500 of Figures 5A through 5E, etc.) to optimize the delivery of images (e.g., first graphical data 108-1 of Figure 2A, graphical data 108-1 of block 404 of method 400 of Figures 4A through 4E, graphical data 108-1 of block 502 of method 500 of Figures 5A through 5E, etc.) over the internet (e.g., communication network 186 of Figure 2A, communication network 186 of Figure 3, communication network 186 of block 422 of method 400 of Figures 4A through 4E, communication network 186 of block 558 of method 500 of Figures 5A through 5E, etc.). Rather than waiting for a high-resolution image to load (e.g., fourth scan 1010-4 of Figure 11), a low-resolution image (e.g., second scan 1010-2 of Figure 11) was displayed immediately and gradually increased in resolution over time. Progressive encoding 126 was performed by encoding a medical image (e.g., graphical data 108) as a series of lossy compressed decompositions (e.g., a plurality of sequence scans 1010), where each subsequent sequence scan 1010 had an increasing level of quality and resolution, such that the complete reconstruction was lossless at the final terminal scan 0110. A progressively encoded byte stream 114 of graphical data 108, with ^^^^ decomposition levels, enabled the reconstruction of ^^^^ + 1 scan 1010, where the resolution of each scan 1010 as a factor of 2 ^^^^+1 of the original full resolution of the graphical data 108. For example, a chest x-ray (“CXR”) with a resolution of 512x512 was progressively encoded 126 with three scan 1010 levels (e.g., three sequence scans in the plurality of sequence scans 1010) and then decoded at sub-resolutions 64x64 for an initial terminal scan, 128x128 for a DB1/ 147040794.2 95 Attorney Ref. No.: 115834-5037-WO second scan, and 256x256 for a third scan as well as at the original resolution 512x512 in the form of the final terminal scan (Figure 14). In some embodiments, the chest x-ray (“CXR”) with a resolution of 512x512 was progressively encoded 126 with three scan 1010 levels then decoded at sub-resolutions 64x64 for an initial terminal scan, 128x128 for a second scan, and 256x256 for the final terminal scan as well as at the original resolution 512x512. This enabled progressively encoded bye streams 114 of graphical data 108 to be decoded using a decoder 128 at different resolutions from a single high-resolution copy. [00378] Moreover, progressively encoded images were decoded with partial byte streams 114. While the JPEG 2000 standard 124 does not explicitly support decoding of partial byte streams 114, this Example allowed for this ability derived from encoding images with tile- part divisions by resolution (e.g., forming the byte stream map 600, block 518 of method 500, etc.). Provided that the bytes to reconstruct the image at a sub-resolution were available, the image was decoded at the decomposition level and all preceding levels even in the absence of the complete byte stream. In contrast, a vast majority of compression codecs (e.g., JPEG 124-1, PNG 124-2) encoded each image sequentially – from the top-left pixel to the bottom- right pixel – which prevented images from being decoded from a partial byte stream in the same way progressive encoding allowed (Figures 15A-B). This enabled quick access to smaller resolution versions of an image (e.g., for thumbnails), thus minimizing data duplication. Despite its tremendous advantages in medical imaging, the computational overhead of progressive encoding has held back its adoption beyond niche web-based applications. [00379] High-Throughput JPEG 2000 (HTJ2K) [00380] Recently, increased popularity of High-Throughput JPEG 2000 (HTJ2K, second code 124-2 of Figure 2B, second code 124-2 of Figure 3, etc.) renewed interest in progressive encoding within the medical imaging community with its integration into the DICOM transfer syntax and cloud-based imaging infrastructures like Amazon Web Services’ HealthImaging. HTJ2K is an extension of the JPEG 2000 standard with significantly faster encoding and decoding at the expense of a slight reduction in encoding efficiency. For lossless encoding and decoding, HTJ2K was more than 30 times faster than JPEG 2000 – putting it on par with the heavily optimized JPEG standard in both single and multi-threaded applications. While compatible with both codecs, this Example utilized HTJ2K for lossless progressive encoding of medical imaging datasets. DB1/ 147040794.2 96 Attorney Ref. No.: 115834-5037-WO [00381] Example 4 – Development of an Intelligent Streaming Framework for Reconstruction of Medical Images at Target Resolution [00382] Intelligent Streaming [00383] This Example discloses the intelligent streaming and subsequent decoding of partial byte streams 114 to reconstruct progressively encoded medical images 108 at an optimal resolution associated with a scan 1010 for AI model 118 inference (e.g., graphical data system 200 of Figure 1, graphical 200 of Figures 2A and 2B, client device 300 of Figure 1, client device 300 of Figure 3, second computer system block 428 of Figure 4C, second computer system of block 532 of method 500, etc.). The optimal resolution (e.g., first characteristic 112-1 of Figure 2A, first training resolution of block 456 of method 400 of Figures 4A-4E, first resolution of block 548 of method 500 of Figures 5A-5E, etc.) was matched based on the smallest decomposition level (e.g., sequence scan 1010) at which the reconstructed medical image 108 contained sufficient information for an AI system (e.g., first model 118-1 of Figure 2A, first model 118-1 of Figure 3, trained feature extraction model of block 418 of method 400 of Figures 4A-4E, trained feature extraction model of block 532 of method 500 of Figures 5A-5E, etc.) to accurately characterize the image (e.g., block 532 of Figure 5C, block 456 of Figure 4E, etc.). An automated system that intelligently determined and validated the optimal resolution associated with a scan 1010 for streaming medical images 108 to an AI system 118 for inference without impacting its diagnostic performance (Figure 16), which comprised of three primary components: a progressive encoder, a stream optimizer, and a progressive decoder. [00384] Progressive Encoder [00385] A progressive encoder as described in Example 3 was obtained. The role of the progressive encoder (e.g., first encoder 126-1 of Figure 2B, first encoder 126-1 of Figure 3, first codec 124-1 of block 422 of method 400 of Figures 4A through 4E, first codec 124-1 of block 558 of method 500 of Figures 5A through 5E, etc.) was to encode medical images into byte streams using HJT2K (Figure 16). Specifically, OpenJPHpy (version 0.1), a Python library for OpenJPH, an open-source implementation of HTJ2K, was utilized to progressively encode imaging data 108 as lossless up to 16-bit depth with 64x64 block size, ^^^^ decomposition levels (e.g., N scans in the plurality of scans), tile-part marker divisions by scan, and tile-part markers to identify the location of the scans within the byte stream 114. The inclusion of tile-part divisions by scan 1010 resolution allowed for decoding of partial DB1/ 147040794.2 97 Attorney Ref. No.: 115834-5037-WO byte stream 114 at different scans 1010 for sub-resolutions. The number of scans 1010 ^^^^ was calculated for each medical image 108 separately prior to encoding 124, which ensured that the sub-resolution of a scan 1010 remained consistent across medical images with different resolutions as follows:
Figure imgf000100_0001
[00387] where ^^^^ and ^^^^ were the horizontal and vertical resolution of the medical image respectively, and ^^^^ was the lower bound for the resolution of the initial terminal scan. ^^^^ = 64 in this Example was set since resolutions smaller than 64x64 have drastically poorer performance in AI systems 118. [00388] The progressive encoder 126 was capable of encoding both 2D and 3D medical images 108. While the encoding of 2D imaging data 108 followed the above methods, the encoding of 3D imaging data 108 required each slice of the volume to be encoded separately as a 2D image. This was particularly applicable to encoding medical images 108 stored in the NIfTI format 124 – which encoded the entire volume in a single file. Furthermore, the rescale intercept and slope were calculated to rescale pixel values to 16-bit unsigned integers when required prior to encoding a medical image (e.g., when encoding Hounsfield units for CT scans). [00389] Stream Optimizer [00390] A stream optimizer was obtained that enabled the streaming and subsequent decoding of partial byte streams 114 to reconstruct medical images at the scan 1010 corresponding to the optimal resolution for the AI system 118. The goal of the stream optimizer was to provide an interface for an AI system 300 to access the database of medical images 106 for inference from the healthcare provider 200. The stream optimizer included two components: a host-side stream optimizer, and a client-side stream optimizer (Figure 16). [00391] The role of the host-side stream optimizer was to interface with the database of progressively encoded medical images (e.g., graphical data store 106 of Figure 2A, etc.) obtained using the progressive encoder described herein and in Example 3, to build a stream optimization map (e.g., byte stream map of 518 of method 500 of Figures 5A-5E, etc.). Each medical image 108 included a unique entry in the stream byte map 600. The stream by ten map 600 allowed for mapping each sub-resolution to the partial byte stream necessary to decode the image at the sub-resolution’s corresponding decomposition level. An entry was DB1/ 147040794.2 98 Attorney Ref. No.: 115834-5037-WO created by locating tile-part markers in the complete byte stream (identified by bytes 0xFF90 and 0xFFD9) and calculating the sub-resolution as a factor of 2 ^^^^+1 of the image’s full resolution. The inclusion of tile-part markers in the encoding step allowed for building of the stream optimization map. This enabled a medical image to be streamed at a specific decomposition level using partial byte streams without streaming the complete byte stream. For example, a CXR with resolution 512x512 was streamed at sub-resolution 256x256 with the first 12 KB of the complete 180 KB byte stream. [00392] On the contrary, the role of the client-side stream optimizer was to interface with the AI system to intelligently determine and validate the optimal resolution for streaming medical images from the host without impacting the AI system’s diagnostic performance. The client- side stream optimizer used the AI vendor’s internal test set to determine the optimal resolution for the AI system. The internal test set was first progressively encoded, then decoded at each decomposition level, and finally evaluated with the AI system to measure performance metrics (e.g., AUROC for classification). The optimal resolution was determined as the smallest decomposition level which corresponded to a sub-resolution closest to the input resolution of the AI system and resulted in no statistically significant difference in the AI system’s performance compared to the full resolution test set. For example, when considering a CXR with resolution 512x512 and deep learning (DL) classification model with input resolution 224x224, the optimal resolution was determined as 256x256 and corresponded to the third decomposition level. [00393] Progressive Decoder [00394] A progressive decoder was further obtained, the role of which was to decode partial byte streams at the specified decomposition level to reconstruct a medical image at the optimal resolution for the AI system (Figure 16). While the decoding of 2D imaging data was straightforward, decoding 3D imaging data required an additional overhead to reconstruct and restitch the volume from the series of 2D slices. For medical images originally in the NIfTI format, each slice of the volume was decoded at the optimal resolution and then stitched into a volume. Then, the original affine transformation matrix and voxel size was rescaled to match the sub-resolution using NiBabel (version 5.2), which ensured correct transformation from voxel coordinates to world coordinates. [00395] Example 5 – Generating Datasets for use in Intelligent Streaming Framework DB1/ 147040794.2 99 Attorney Ref. No.: 115834-5037-WO [00396] In this Example, six medical imaging datasets were obtained, each dataset including graphical data in the form of medical images. Three datasets were obtained for each of (i) 2D classification with CXRs and (ii) 3D segmentation with CT scans. [00397] Datasets for Classification Task [00398] The National Institutes of Health’s (NIH) ChestX-ray14 dataset (hereafter, ‘NIH’) included of n=112,120 frontal CXRs from 30,805 patients. All images in the dataset had a fixed resolution of 1024x1024, had pixel values normalized to range 0–255, and were encoded in the PNG format (e.g., lossless). Seven disease labels from the NIH dataset were considered for analysis: Atelectasis, Cardiomegaly, Consolidation, Edema, Pleural Effusion, Pneumonia, and Pneumothorax. [00399] The CheXpert dataset included 224,316 frontal and lateral CXRs from 65,240 patients. Images in the dataset had a mean resolution of 2828x2320, had pixel values normalized to range 0–255, and were encoded in the JPEG format (e.g., lossy). For inference, n=15,000 frontal CXRs were randomly sampled and all uncertain labels were treated as negatives. The same seven disease labels as the NIH dataset were considered for analysis. [00400] The MIMIC-CXR-JPG dataset (hereafter, ‘MIMIC’) included n=377,110 frontal and lateral CXRs from 65,379 patients. Images in the dataset had a mean resolution of 2500x3056, had pixel values normalized with histogram equalization to range 0–255, and were encoded in the JPEG format (e.g., lossy). Furthermore, the MIMIC-CXR dataset provides the images in the 12-bit uncompressed DICOM Little Endian format. The same preprocessing steps, excluding JPEG conversion, were applied to the DICOM images during inference. For inference, n=15,000 frontal CXRs were randomly sampled and all uncertain labels were treated as negatives. The same seven disease labels as the NIH dataset were considered for analysis. [00401] Datasets for Segmentation Task [00402] The Medical Segmentation Decathlon (MSD) is a collection of benchmark datasets for 3D segmentation including 10 tasks spanning across different body parts and imaging modalities. Out of the MSD collection, two datasets were utilized: 1) The MSD Liver dataset included n=131 abdomen portal venous phase contrast-enhanced CT scans containing liver and liver tumor annotations, in which all tumor annotations were discarded, and the MSD Spleen dataset included n=41 abdomen non-enhanced CT scans with spleen annotations. DB1/ 147040794.2 100 Attorney Ref. No.: 115834-5037-WO Both datasets included scans acquired at standard CT axial resolution of 512x512 with variable volume sizes and encoded in the NIfTI format (e.g., lossless). [00403] The Multi-Atlas Labeling Beyond the Cranial Vault (BTCV) dataset included n=30 abdomen portal venous phase contrast-enhanced CT scans with annotations for 13 abdominal organs. The dataset included scans acquired at standard CT axial resolution of 512x512 with variable volume sizes and encoded in the NIfTI format (e.g., lossless). Annotations for the liver and spleen organs were utilized to create an external test set for each organ. [00404] Example 6 – Assay Development for Classification and Segmentation of Graphical Data using an Intelligent Streaming Framework [00405] Experimental Design [00406] AI inference 118 was evaluated across two experiments: 2D classification of abnormalities with CXRs, and 3D segmentation of the liver and spleen organs with abdomen CT scans using the datasets for each task described in Example 5. In both cases, a deep learning (DL) model 118 (in accordance with the framework described in Examples 3 and 4) was trained using an internal dataset (e.g., a training data set comprising a first plurality of graphical data, in accordance with the present disclosure), the optimal resolution for AI inference was determined and validated using an internal test set from the same site, and the optimal resolution was then evaluated on external data streamed from a different site (e.g., a second one or more graphical data, in accordance with the present disclosure). [00407] For each test set, the mean structural similarity index measure (SSIM) and mean peak signal-to-noise ratio (PSNR) were measured as image fidelity metrics. The amount of transferred data, decoding time, inference time, and throughput were measured as data efficiency metrics. Both sets of metrics were measured at each decomposition level (including the optimal resolution) for the Intelligent Streaming framework (“ISLE”) streamed version and compared with the original non-ISLE streamed versions of the test set. The SSIM and PSNR after normalizing pixel values to range 0–1 were measured using scikit- image (version 0.22). The decode time was measured using OpenCV (version 4.7) for JPEG and PNG formats, Pydicom (version 2.3.1) for uncompressed DICOM imaging data, and OpenJPHpy (version 0.1) for HTJ2K format. The inference time was measured by the time taken by the AI system to characterize a medical image, including any preprocessing steps such as resizing to input resolution. Both decoding and inference times were measured on a DB1/ 147040794.2 101 Attorney Ref. No.: 115834-5037-WO single thread, without parallel processing. The throughput of the AI system was measured as the number of scans processed by the AI system per second and is defined as follows: ^^^^ℎ ^^^^ ^^^^ ^^^^ ^^^^ℎ ^^^^ ^^^^ ^^^^ = ^^^^ ^^^^. ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ + ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ ^^^^ [00408] Classification Task [00409] The internal NIH dataset described in Example 5 was randomly split into training (70%, n=78,075), validation (10%, n=11,079), and testing (20%, n=22,966) sets with no patient leakage. Then, a multi-label DenseNet121 trained model 118, pre-trained with ImageNet, using transfer learning for 100 epochs with binary cross-entropy loss, batch size of 64, learning rate of 5e-5, and early stopping. All CXRs were resized to resolution 224x224 while maintaining aspect ratio with zero-padding with pixel values normalized to range 0–1 and scaled to ImageNet statistics. Random augmentations (e.g., rotation, flip, zoom, contrast, or a combination thereof) were applied to input images during training. Then, the internal NIH test set was used to determine and validate the optimal resolution. Finally, the optimal resolution on the external CheXpert and MIMIC-CXR-JPG datasets was evaluated. All models 118 were trained and tested using TensorFlow (version 2.8.1) and CUDA (version 12.0) on a NVIDIA GeForce RTX 4090 GPU. Model 118 performance was measured by the mean area under the receiver operating characteristic curve (AUROC) scores of the seven disease labels. [00410] Segmentation Task [00411] For liver segmentation, the internal MSD Liver dataset described in Example 5 was randomly into training (80%, n=106) and testing (20%, n=25) sets. Then, a modified 3D- UNet model 118 was trained with five levels of two residual convolutional units on the internal MSD Liver dataset for 500 epochs with Dice loss, batch size of 2, decaying learning rate with cosine annealing scheduler, and initial learning rate of 1e-4. All CT scans were resized to resolution 256x256x128, voxel spacing rescaled to 1.5x1.5x2.0 mm, and pixel values normalized to range 0–1. Random foreground patches of resolution 128x128x32 were extracted from each volume and centered on voxels belonging either to the foreground or background class. Random augmentations (e.g., rotation, scaling, elastic deformation, or a combination thereof) were applied to input volumes during training. Then, the internal MSD Liver test set was used to determine and validate the optimal resolution using our framework. Finally, the optimal resolution was determined on the external BTCV Liver dataset. DB1/ 147040794.2 102 Attorney Ref. No.: 115834-5037-WO [00412] For spleen segmentation, the internal MSD Spleen dataset described in Example 5 was randomly split into training (80%, n=32) and testing (20%, n=9) sets. Then, similarly trained a modified 3D-UNet model using the methods described supra. Then, the internal MSD Spleen test set was used to determine and validate the optimal resolution using t the encoded byte stream 114. Finally, the optimal resolution on the external BTCV Spleen dataset was determined. All models 118 were trained and tested using PyTorch (version 1.13), MONAI (version 1.1), and CUDA (version 12.0) on a NVIDIA GeForce RTX 4090 GPU. Model performance was measured by the mean Dice similarity coefficient. [00413] Statistical Analysis [00414] Performance metrics using statistical tests between the original dataset and each decomposition were compared for each test set. For classification, the mean AUROC scores were compared with a one-sided paired t-test due to the normal distribution of paired samples, indicated by P > 0.05 using the Shapiro-Wilk test for normality. For segmentation, the mean Dice scores were compared with a one-sided Wilcoxon rank-sum test due to the non-parametric distribution of paired samples. In both cases, the alternative hypothesis that the mean performance metric for a decomposition is less than the mean performance metric for the original dataset was tested. Statistical significance was defined as P ≤ 0.05. [00415] Example 7 – Intelligent Streaming Framework Optimizes Decoding of Graphical Data for Classification Tasks by Determining Optimal Resolution of Images [00416] Results – Classification [00417] Using the internal NIH test set described in Example 5 and according to the assay parameters described in Example 6, the optimal resolution for the DL model was determined as the third decomposition level with resolution 256x256, which was the closest resolution to the model’s input size of 224x224. The optimal resolution measured a mean AUROC of 0.84 (SD = 0.05) compared to the mean AUROC of 0.84 (SD = 0.05) with the original PNG data (P = 0.62). At the optimal resolution, the encoded byte stream 114 required 91.58% less data to be transferred, 87.65% less time to decode, and 11.61% less time for inference while increasing throughput by 3.72x. Furthermore, the optimal resolution measured a mean SSIM of 0.97 (SD = 0.01) and mean PSNR of 31.51 (SD = 3.96). The results are detailed in Table 1, below, and Figure 17A. Table 1: Data efficiency and performance metrics on the internal NIH test set (n=22,966) for each decomposition level and the original format. The optimal resolution appeared at DB1/ 147040794.2 103 Attorney Ref. No.: 115834-5037-WO decomposition 3, in bold below. The mean AUROC scores were reported in Mean (SD) format. Performance metrics were compared using one-sided paired t-tests. Statistically significant comparisons are underlined below. Format Decomp. 124 (Scan Data Transfer Decode Inference Throughput Mean P-value 101 (GB) Time (s) Time (s) (scans/s) AUROC
Figure imgf000106_0001
2 0.23 6.20 13.94 1140.11 0.836 (0.050) 0.07 3 0.72 15.03 14.12 787.78 0.838 (0.050) 0.62 4 2.50 40.09 14.63 419.68 0.838 (0.052) 0.72 5 8.26 120.50 15.26 169.17 0.838 (0.052) 1.00 [00418] When evaluating on the external CheXpert test set, the optimal resolution measured a mean AUROC of 0.71 (SD = 0.08) compared to the mean AUROC of 0.70 (SD = 0.08) with the original JPEG data (P = 0.99). However, the ISLE framework required 97.43% less data to be transferred, 97.70% less time to decode, and 13.46% less time for inference while increasing throughput by 23.69x. Furthermore, the optimal resolution measured a mean SSIM of 0.92 (SD = 0.03) and mean PSNR of 28.57 (SD = 1.70). The results are detailed in Table 2, below, and Figure 17B. Table 2: Data efficiency and performance metrics on the external CheXpert test set (n=15,000) for each decomposition level and the original format. The optimal resolution appeared at decomposition 3, in bold below. The mean AUROC scores were reported in Mean (SD) format. Performance metrics were compared using one-sided paired t-tests. Statistically significant comparisons are underlined below. Format Decomp. (Sca Data Transfer Decode Inference Throughput Mean 124 n (GB) Time (s) Time (s) (scans/ P-value 101 s) AUROC
Figure imgf000106_0002
JPEG 535.69 11.35 27.42 0.699
Figure imgf000106_0006
(0.0 -
Figure imgf000106_0003
Figure imgf000106_0004
Figure imgf000106_0005
78) HTJ2K 1 0.09 4.87 9.02 1079.33 0.698 (0.084) 0.42 2 0.24 8.53 9.50 831.92 0.713 (0.084) 0.99 DB1/ 147040794.2 104 Attorney Ref. No.: 115834-5037-WO 3 0.81 12.34 9.82 676.99 0.709 (0.082) 0.99 4 3.08 29.44 9.85 381.78 0.706 (0.079) 0.99 5 12.11 93.80 9.55 145.14
Figure imgf000107_0001
(0.079) 0.99 6 46.04 356.85 10.96 40.78 0.698 (0.079) 0.04 7 52.08 402.42 11.22 36.26 0.699 (0.078) 1.00 [00419] Similarly, when evaluating on the external MIMIC dataset, the optimal resolution measured a mean AUROC of 0.77 (SD = 0.07) compared to the mean AUROC of 0.78 (SD = 0.07) with the uncompressed DICOM data (P = 0.99). More importantly, ISLE required 99.83% less data to be transferred, 99.53% less time to decode, and 97.23% less time for inference while increasing the throughput by 113.87x. When comparing the ISLE streamed dataset with the JPEG version of the MIMIC dataset, no significant differences in the mean AUROC of 0.79 (SD = 0.07, P = 1.00) were observed. The ISLE framework only required 98.62% less data to be transmitted, 98.49% less time to decode, and 18.73% less time for inference while increasing throughput by 30.12x. Furthermore, the optimal resolution measured a mean SSIM of 0.92 (SD = 0.03) and mean PSNR of 27.30 (SD = 2.34). The results are detailed in Table 3, below, and Figure 17C. Table 3: Data efficiency and performance metrics on the external MIMIC test set (n=15,000) for each decomposition level and the original format. Both sets of metrics are also provided for the uncompressed DICOM version of the MIMIC dataset. The optimal resolution appeared at decomposition 3, in bold below. The mean AUROC scores were reported in Mean (SD) format. Performance metrics were compared using one-sided paired t-tests. Statistically significant comparisons are underlined below. Format Decomp. Data Transfer Decode Infere 124 (Scan nce Throughput Mean (GB) Time (s) Time (s) (scans/s) AUROC P-value
Figure imgf000107_0002
DICOM - 199.24 1540.88 10.46 9.67 0.779 (0.068) 1.00 HTJ2K 1 0.04 2.90 8.52 1314.07 0.722 (0.076) 0.002 2 0.11 4.00 8.55 1195.22 0.767 (0.072) 0.88 DB1/ 147040794.2 105 Attorney Ref. No.: 115834-5037-WO 3 0.34 7.44 8.71 928.76 4 1.24 17.03 8.76 581.53 5 4.70 47.91 8.98 263.66 6 17.54 161.99 9.54 87.45 7 43.94 438.59 10.20 33.42
Figure imgf000108_0001
[00420] Example 8 – Intelligent Streaming Framework Optimizes Decoding of Graphical Data for Segmentation Tasks by Determining Optimal Resolution of Images [00421] Results – Segmentation [00422] Liver Segmentation [00423] Using the internal MSD Liver test set described in Example 5 and according to the assay parameters described in Example 6, the ISLE framework determined the optimal resolution of the DL model as the third decomposition with resolution 256x256 and variable volumes sizes, which was the closest resolution to the model’s input size of 256x256x128. The optimal resolution measured a mean Dice of 0.95 (SD = 0.02) compared to the mean Dice of 0.95 (SD = 0.02) with the original NIfTI data (P = 0.27). At the optimal resolution, the ISLE framework required 81.58% less data to be transferred, 87.86% less time to decode, and 2.03% less time for inference while increasing the throughput by 3.42x. Furthermore, the optimal resolution measured a mean SSIM of 0.96 (SD = 0.02) and mean PSNR of 33.50 (SD = 5.76). The results are detailed in Table 4, below, and Figure 18A. Table 4: Data efficiency and performance metrics on the internal MSD Liver test set (n=25) for each decomposition level and the original format. The optimal resolution appeared at decomposition 3, in bold below. The mean Dice scores were reported in Mean (SD) format. Performance metrics were compared using one-sided Wilcoxon rank-sum tests. Statistically significant comparisons are underlined below. Format Decomp. Data Transfer Decode Inference Throughput 124 (Scan Mean (GB) Time (s) Time (s) (scans/s) Dice P-value NIfTI 4.90 75.51 10.51 0.35 0.945
Figure imgf000108_0002
(0.017) -
Figure imgf000108_0003
Figure imgf000108_0004
Figure imgf000108_0005
Figure imgf000108_0006
HTJ2K 1 0.13 2.40 10.36 2.35 0.917 (0.021) ≤ 0.001 DB1/ 147040794.2 106 Attorney Ref. No.: 115834-5037-WO 2 0.30 3.98 10.29 2.10 0.937 (0.021) ≤ 0.001 3 0.90 9.17 10.30 1.54 0.945 (0.019) 0.27 4 2.88 29.24 10.37 0.76 0.945 (0.017) 1.00 [00424] When evaluating on the external BTCV dataset, the optimal resolution measured a mean Dice of 0.93 (SD = 0.02) compared to the mean Dice of 0.93 (SD = 0.02) on the original NIfTI data (P = 1.00). More importantly, the ISLE framework required 77.93% less data to be transferred, 89.50% less time to decode, and 1.34% less time for inference while increasing throughput by 2.60x. Furthermore, the optimal resolution measured a mean SSIM of 0.97 (SD = 0.00) and mean PSNR of 30.80 (SD = 2.80). The results are detailed in Table 5, below, and Figure 18A. Table 5: Data efficiency and performance metrics on the external BTCV Liver test set (n=30) for each decomposition level and the original format. The optimal resolution appeared at decomposition 3, in bold below. The mean Dice scores were reported in Mean (SD) format. Performance metrics were compared using one-sided Wilcoxon rank-sum tests. Statistically significant comparisons are underlined below. Format Decomp. Data Transfer Decode Inference Throughput Mea 124 (Scan n (GB) Time (s) Time (s) (scans/s) D P-value 1010 ice
Figure imgf000109_0001
[00425] Spleen Segmentation [00426] Using the internal MSD Spleen test set described in Example 5 and according to the assay parameters described in Example 6, the ISLE framework similarly determined the optimal resolution of the DL model as the third decomposition with resolution 256x256 and variable volumes sizes due to the shared architecture of both DL models. The optimal DB1/ 147040794.2 107 Attorney Ref. No.: 115834-5037-WO resolution measured a mean Dice of 0.93 (SD = 0.02) compared to the mean Dice of 0.93 (SD = 0.02) with the original NIfTI data (P = 0.37). At the optimal resolution, the ISLE framework required 81.99% less data to be transferred, 93.27% less time to decode, and 3.92% less time for inference while increasing the throughput by 2.87x. Furthermore, the optimal resolution measured a mean SSIM of 0.96 (SD = 0.01) and mean PSNR of 30.20 (SD = 0.93). The results are detailed in Table 6, below, and Figure 18B. Table 6: Data efficiency and performance metrics on the internal MSD Spleen test set (n=9) for each decomposition level and the original format. The optimal resolution appeared at decomposition 3, in bold below. The mean Dice scores were reported in Mean (SD) format. Performance metrics were compared using one-sided Wilcoxon rank-sum tests. Statistically significant comparisons are underlined below. Format Decomp. Data Transfer Decode Inference Throughput Mean 124 (Scan (GB) Time (s) Time (s) (scans/s) Dice P-value
Figure imgf000110_0001
3 0.03 0.43 1.68 14.24 0.934 (0.020) 0.37 4 0.10 1.56 1.53 9.72 0.934 (0.021) 1.00 [00427] When evaluating on the external BTCV dataset, the optimal resolution measured a mean Dice of 0.88 (SD = 0.14) compared to the mean Dice of 0.88 (SD = 0.15) on the original NIfTI data (P = 0.97). The ISLE framework required 77.93% less data to be transferred, 89.50% less time to decode, and 3.24% less time for inference while increasing throughput by 2.62x. Furthermore, the optimal resolution measured a mean SSIM of 0.97 (SD = 0.00) and mean PSNR of 30.80 (SD = 2.80). The results are detailed in Table 7, below, and Figure 18B. [00428] Table 7: Data efficiency and performance metrics on the external BTCV Spleen test set (n=30) for each decomposition level and the original format. The optimal resolution appeared at decomposition 3, in bold below. The mean Dice scores were reported in Mean (SD) format. Performance metrics were compared using one-sided Wilcoxon rank-sum tests. Statistically significant comparisons are underlined below. DB1/ 147040794.2 108 Attorney Ref. No.: 115834-5037-WO Format Decomp. (Scan Data Transfer Decode Inference Throughput Mean 124 (GB) Time (s) Time (s) (scans/s) Dice P-value
Figure imgf000111_0001
3 0.20 2.35 5.38 3.88 0.881 (0.143) 0.97 4 0.63 6.66 5.49 2.47 0.879 (0.147) 1.00 [00429] The detailed breakdown of the encoding time to progressively encode all six medical imaging datasets is provided in Table 8, below. Table 8: Encoding time for all six medical imaging datasets considered in this study measured. All encoding time is measured on a single thread, without parallel processing. The encoding time per scan is reported in the Mean (SD) format. Dataset Format 124 No. of Encoding Time Per Total Encoding Scans 1010 Scan (ms) Time (s) NIH ChestX-Ray14
Figure imgf000111_0003
22,966
Figure imgf000111_0004
11.87 (1.79)
Figure imgf000111_0002
272.60
Figure imgf000111_0005
CheXpert JPEG 15,000 79.10 (28.58) 1186.43 MIMIC-CXR-JPG JPEG 15,000 76.22 (21.13) 1143.28 MSD Liver NIfTI 25 2565.91 (867.20) 64.15 MSD Spleen NIfTI 9 211.65 (104.24) 1.90 BTCV NIfTI 30 403.37 (111.13) 12.10 [00430] Examples of CXRs and abdomen CT scans at various decomposition levels are provided in Figures 19 and 20. [00431] The image fidelity metrics for each decomposition level are detailed in Tables 9-14, below, which provides a detailed breakdown of image fidelity metrics for all six medical images datasets. The optimal resolution is highlighted in bold, below. The SSIM and PSNR values were reported in Mean (SD) format. A PSNR of infinity (abbreviated, ‘Inf’) below implies a perfect reconstruction of the image. Table 9: NIH Dataset DB1/ 147040794.2 109 Attorney Ref. No.: 115834-5037-WO Format Decomposition SSIM PSNR
Figure imgf000112_0001
HTJ2K 1 0.82 (0.05) 24.80 (1.73) 2 0.87 (0.03) 26.24 (1.49) 3 0.92 (0.03) 28.57 (1.70) 4 0.94 (0.03) 31.07 (1.85) 5 0.97 (0.02) Inf Table 10: CheXpert Dataset Format Decomposition SSIM PSNR
Figure imgf000112_0002
HTJ2K 1 0.82 (0.05) 24.80 (1.73) 2 0.87 (0.03) 26.24 (1.49) 3 0.92 (0.03) 28.57 (1.70) 4 0.94 (0.03) 31.07 (1.85) 5 0.97 (0.02) Inf 6 1.00 (0.01) Inf 7 1.00 (0.00) Inf Table 11: MIMIC Dataset Format Decomposition SSIM PSNR
Figure imgf000112_0003
DICOM - 0.88 (0.07) 18.69 (3.35) HTJ2K 1 0.82 (0.06) 23.30 (2.22) 2 0.90 (0.03) 26.00 (2.05) 3 0.92 (0.03) 27.30 (2.34) 4 0.95 (0.02) 30.21 (2.63) 5 0.97 (0.01) Inf 6 0.99 (0.01) Inf 7 1.00 (0.00) Inf Table 12: MSD Liver Dataset Format Decomposition SSIM PSNR
Figure imgf000112_0004
DB1/ 147040794.2 110 Attorney Ref. No.: 115834-5037-WO HTJ2K 1 0.72 (0.09) 23.41 (4.70) 2 0.85 (0.06) 26.94 (5.46) 3 0.96 (0.02) 33.50 (5.76) 4 1.00 (0.00) Inf Table 13: MSD Spleen Dataset Format Decomposition SSIM PSNR
Figure imgf000113_0001
HTJ2K 1 0.68 (0.03) 20.04 (0.94) 2 0.84 (0.03) 23.94 (2.03) 3 0.96 (0.01) 30.20 (0.93) 4 1.00 (0.00) Inf Table 14: BTCV Dataset Format Decomposition SSIM PSNR
Figure imgf000113_0002
HTJ2K 1 0.76 (0.05) 21.93 (2.22) 2 0.89 (0.02) 25.96 (1.73) 3 0.97 (0.00) 30.80 (2.80) 4 1.00 (0.00) Inf [00432] Discussion [00433] Examples 3-8 demonstrated the potential of the present disclosure to dramatically reduce the amount of data transmitted and decoding time when streaming medical images over the communication network, while significantly increasing the throughput of AI systems for inference across classification and segmentation tasks. For all three tasks, the optimal resolution determined on the internal test set scaled well to external data and did not impact the performance of the DL model. This was expected as inference on a medical image with resolution close to the input size of a DL model should not impact the diagnostic performance of the AI system. However, the example indicated that segmentation tasks are more sensitive to lower resolution images and lossy compression artifacts than classification tasks – a finding that aligned with prior literature. DB1/ 147040794.2 111 Attorney Ref. No.: 115834-5037-WO [00434] The primary advantage of the encoded byte stream 114 was the ability to intelligently stream medical images at the optimal resolution for inference to an AI vendor without additional image preprocessing or data duplication. This enabled healthcare providers to store medical images at a single high-resolution copy while streaming to different AI vendors at different resolutions. While compression was widely adopted in imaging infrastructures for reducing data storage and transmission, the encoded byte stream 114 takes further enabled healthcare providers and AI vendors to benefit from the inherent gains in data efficiency from streaming a lower resolution image. For example, using the encoded byte stream 114, the uncompressed external MIMIC test set shrunk almost 600 times in size – from 199 GB to 0.34 GB – when the medical images were progressively encoded using HTJ2K and then streamed at the optimal resolution. [00435] The streaming of medical images at the optimal resolution also improved the computational efficiency of AI inference – a key factor for reducing the carbon footprint of AI deployments. A vast majority of DL models 118 required a fixed input resolution. Therefore, medical images with resolutions larger than the input resolution were resized prior to inference. This bottleneck resulted in diminishing returns in net image fidelity and DL model performance with increasing resolution. While the computational efficiency of popular resizing algorithms (e.g., bilinear interpolation) was independent of the original image resolution, the computational efficiency of decoding and loading medical images into memory for inference was dependent on it. Examples 3-8 illustrate that the encoded byte stream 114 reduced the decoding and inference time, resulting in a dramatic increase in the throughput of AI systems. [00436] As the adoption of AI systems grow within the clinical environment, smaller practices and rural clinics are not able to keep pace with larger hospital systems in terms of implementing high-quality computational infrastructure and, therefore, are not able to support the deployment of large-scale AI solutions, leading to widening of health disparities in the US population. The encoded byte stream 114 advantageously allows for the maintenance of equity in care across both small and large healthcare systems by dramatically reducing the infrastructural requirements for healthcare providers. [00437] Examples 3-8 indicate that the benefits of the encoded byte stream 114 are not just dependent on the task and modality, but also the architecture of the AI system. For example, a DL segmentation model 118 with CT scans trained on standard acquisition resolutions yielded significantly fewer gains in data and computational efficiency compared to a model DB1/ 147040794.2 112 Attorney Ref. No.: 115834-5037-WO trained on lower resolutions. Furthermore, due to the limitations of HTJ2K, the encoded byte stream 114 did not perform 3D compression and each axial slice of the volume was encoded and decoded as a separate image. This added an additional step to reconstruct the volume after decoding. Finally, the encoded byte stream 114 utilized medical images progressively encoded using either JPEG 2000 or HTJ2K for intelligent streaming. This was allowed existing imaging infrastructures that do not use compression to store medical images. [00438] Examples 3-8 explored training DL models such that they were robust on inference with low resolution images for medical image AI applications on the edge through techniques like data augmentation. These Examples explored the potential for reducing the amount of data storage and transmission for cloud-based imaging databases with a focus on greater accessibility and interoperability of large-scale medical images datasets for AI applications. [00439] These Examples presented an important first step towards a framework for intelligently streaming medical images to AI vendors at an optimal resolution for inference. The encoded byte stream 114 addressed inefficiencies in current imaging infrastructures by improving data and computational efficiency of AI deployments in the clinical environment without impacting clinical decision-making using AI systems. [00440] Additional details and information regrading method 400, method 500, and Examples 1-8 is found at Hosny A, Parmar C, Quackenbush J, Schwartz LH, Aerts HJWL. Artificial intelligence in radiology. Nat Rev Cancer.2018;18(8):500-510. doi:10.1038/s41568-018-0016-5; Thrall JH, Li X, Li Q, et al. Artificial Intelligence and Machine Learning in Radiology: Opportunities, Challenges, Pitfalls, and Criteria for Success. J Am Coll Radiol.2018;15(3):504-508. doi:10.1016/j.jacr.2017.12.026; Wong S, Zaremba L, Gooden D, Huang HK. Radiologic image compression-a review. Proc IEEE.1995;83(2):194- 219. doi:10.1109/5.364466; Sabottke CF, Spieler BM. The Effect of Image Resolution on Deep Learning in Radiography. Radiol Artif Intell.2020;2(1):e190015. doi:10.1148/ryai.2019190015; Huda W, Abrahams RB. X-Ray-Based Medical Imaging and Resolution. Am J Roentgenol.2015;204(4):W393-W397. doi:10.2214/AJR.14.13126; Noumeir R, Pambrun JF. Using JPEG 2000 Interactive Protocol to Stream a Large Image or a Large Image Set. J Digit Imaging.2011;24(5):833-843. doi:10.1007/s10278-010-9343-0; Jo YY, Choi YS, Park HW, et al. Impact of image compression on deep learning-based mammogram classification. Sci Rep.2021;11(1):7924. doi:10.1038/s41598-021-86726-w; Shih G, Wu CC, Halabi SS, et al. Augmenting the National Institutes of Health Chest Radiograph Dataset with Expert Annotations of Possible Pneumonia. Radiol Artif Intell. DB1/ 147040794.2 113 Attorney Ref. No.: 115834-5037-WO 2019;1(1):e180041. doi:10.1148/ryai.2019180041; Lehmann TM, Abel J, Weiss C. The impact of lossless image compression to radiographs. In: Medical Imaging 2006: PACS and Imaging Informatics. Vol 6145. ; 2006:290-297. doi:10.1117/12.651697; Clunie DA. Lossless compression of grayscale medical images: effectiveness of traditional and state-of- the-art approaches. In: Medical Imaging 2000: PACS Design and Evaluation: Engineering and Clinical Issues. Vol 3980. ; 2000:74-84. doi:10.1117/12.386389; Koff DA, Shulman H. An overview of digital compression of medical images: can we use lossy image compression in radiology? Can Assoc Radiol J J Assoc Can Radiol.2006;57(4):211-217; Koff D, Bak P, Brownrigg P, et al. Pan-Canadian Evaluation of Irreversible Compression Ratios (“Lossy” Compression) for Development of National Guidelines. J Digit Imaging.2009;22(6):569-578. doi:10.1007/s10278-008-9139-7; Johnson AEW, Pollard TJ, Greenbaum NR, et al. MIMIC- CXR-JPG, a large publicly available database of labeled chest radiographs. Published online November 14, 2019. doi:10.48550/arXiv.1901.07042; Foos DH, Muka E, Slone RM, et al. JPEG 2000 compression of medical imagery. In: Blaine GJ, Siegel EL, eds. Medical Imaging 2000: PACS Design and Evaluation: Engineering and Clinical Issues. Vol 3980. ; 2000:85- 96. doi:10.1117/12.386390; HTJ2K Transfer Syntax. Published online November 14, 2023. Accessed February 21, 2024. dicom.nema.org/medical/dicom/Final/sup235_ft_HTJ2K.pdf; AWS HealthImaging. Accessed February 21, 2024. aws.amazon.com/healthimaging/; High Throughput JPEG 2000 (HTJ2K) and the JPH file format: a primer. Accessed February 21, 2024. ds.jpeg.org/whitepapers/jpeg-htj2k-whitepaper.pdf; Taubman D, Naman A, Mathew R, Smith M, Watanabe O, Lemieux PA. High throughput JPEG 2000 (HTJ2K): Algorithm, performance and potential. Published online May 29, 2020. Accessed February 21, 2024. htj2k.com/wp-content/uploads/white-paper.pdf; Boliek M, Christopoulos C, Majani E. JPEG 2000 Image Coding System. Published online April 11, 2000. Accessed February 26, 2024. ics.uci.edu/~dan/class/267/papers/jpeg2000.pdf; Wang X, Peng Y, Lu L, Lu Z, Bagheri M, Summers RM. ChestX-Ray8: Hospital-Scale Chest X-Ray Database and Benchmarks on Weakly-Supervised Classification and Localization of Common Thorax Diseases. In: 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). IEEE; 2017:3462- 3471. doi:10.1109/CVPR.2017.369; Irvin J, Rajpurkar P, Ko M, et al. CheXpert: A Large Chest Radiograph Dataset with Uncertainty Labels and Expert Comparison. Proc AAAI Conf Artif Intell.2019;33(01):590-597. doi:10.1609/aaai.v33i01.3301590; Garbin C, Rajpurkar P, Irvin J, Lungren MP, Marques O. Structured dataset documentation: a datasheet for CheXpert. Published online May 6, 2021. doi:10.48550/arXiv.2105.03020; Goldberger AL, Amaral LAN, Glass L, et al. PhysioBank, PhysioToolkit, and PhysioNet: Components of a DB1/ 147040794.2 114 Attorney Ref. No.: 115834-5037-WO New Research Resource for Complex Physiologic Signals. Circulation.2000;101(23). doi:10.1161/01.CIR.101.23.e215; Johnson AEW, Pollard TJ, Berkowitz SJ, et al. MIMIC- CXR, a de-identified publicly available database of chest radiographs with free-text reports. Sci Data.2019;6(1):317. doi:10.1038/s41597-019-0322-0; Antonelli M, Reinke A, Bakas S, et al. The Medical Segmentation Decathlon. Nat Commun.2022;13(1):4128. doi:10.1038/s41467-022-30695-9; Landman B, Xu Z, Igelsias J, Styner M, Langerak T, Klein A. MICCAI Multi-Atlas Labeling Beyond the Cranial Vault - Workshop and Challenge. Published online 2015. doi:10.7303/SYN3193805; Liu Z, Zhuang J, Xu X, et al. Machine Vision Guided 3D Medical Image Compression for Efficient Transmission and Accurate Segmentation in the Clouds. In: 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). IEEE; 2019:12679-12688. doi:10.1109/CVPR.2019.01297; Doo FX, Vosshenrich J, Cook TS, et al. Environmental Sustainability and AI in Radiology: A Double-Edged Sword. Radiology.2024;310(2):e232030. doi:10.1148/radiol.232030; Douthit N, Kiv S, Dwolatzky T, Biswas S. Exposing some important barriers to health care access in the rural USA. Public Health.2015;129(6):611-620. doi:10.1016/j.puhe.2015.04.001; Doo FX, Kulkarni P, Siegel EL, et al. Economic and Environmental Costs of Cloud Technologies for Medical Imaging and Radiology Artificial Intelligence. J Am Coll Radiol. 2024;21(2):248-256. doi:10.1016/j.jacr.2023.11.011, each of which is hereby incorporated by reference in its entirety for all purposes. REFERENCES CITED AND ALTERNATIVE EMBODIMENTS [00441] All references cited herein are incorporated herein by reference in their entirety and for all purposes to the same extent as if each individual publication or patent or patent application was specifically and individually indicated to be incorporated by reference in its entirety for all purposes. [00442] The present invention can be implemented as a computer program product that includes a computer program mechanism embedded in a non-transitory computer-readable storage medium. For instance, the computer program product could contain instructions for operating the user interfaces disclosed herein. These program modules can be stored on a CD-ROM, DVD, magnetic disk storage product, USB key, or any other non-transitory computer readable data or program storage product. DB1/ 147040794.2 115 Attorney Ref. No.: 115834-5037-WO [00443] Many modifications and variations of this invention can be made without departing from its spirit and scope, as will be apparent to those skilled in the art. The specific embodiments described herein are offered by way of example only. The embodiments were chosen and described in order to best explain the principles of the invention and its practical applications, to thereby enable others skilled in the art to best utilize the invention and various embodiments with various modifications as are suited to the particular use contemplated. The invention is to be limited only by the terms of the appended claims, along with the full scope of equivalents to which such claims are entitled. DB1/ 147040794.2 116

Claims

Attorney Ref. No.: 115834-5037-WO WHAT IS CLAIMED IS: 1. A method for optimizing decoding of graphical data, the method comprising: at a first computer system comprising a processor and memory: A) obtaining, in electronic form, a training data set comprising a first plurality of graphical data, wherein the first plurality of graphical data is defined by a first plurality of characteristics comprising: a sampling resolution of the first plurality of graphical data, and a first modality associated with a capture of the first plurality of graphical data; B) using the first plurality of graphical data to train a feature extraction model, thereby obtaining (i) a trained feature extraction model and (ii) performance data, the performance data comprising a plurality of training resolutions, wherein each respective training resolution in the plurality of training resolutions is less than the sampling resolution and is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances; C) receiving a request to evaluate a second one or more graphical data at a second computer system, wherein the second one or more graphical data is defined by a second plurality of characteristics comprising: the first modality associated with a capture of the second one or more graphical data, and a desired threshold performance; D) encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data, thereby forming an encoded byte stream comprising a plurality of sequence scans through the sampling resolution, wherein: each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans, and each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans; and E) matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data; and DB1/ 147040794.2 117 Attorney Ref. No.: 115834-5037-WO F) communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream at the second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, thereby optimizing decoding of graphical data. 2. The method of claim 1, wherein the first plurality of graphical data comprises one or more images taken of a first biological sample. 3. The method of claim 2, wherein the second one or more graphical data comprises one or more images taken of a second biological sample different than the first biological sample. 4. The method of any preceding claim, wherein the first plurality of graphical data comprises two-dimensional (2D) graphical data, three-dimensional (3D) graphical data, four- dimensional (4D) data, or a combination thereof. 5. The method according to any one of claims 1-3, wherein the second one or more graphical data comprises 3D graphical data and each scan in the plurality of sequence scans of the 3D graphical data comprises a respective 2D layer of the 3D graphical data. 6. The method of any preceding claim, wherein the first plurality of graphical data comprises one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. 7. The method of claim 6, wherein the first plurality of graphical data comprises 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images. 8. The method of any preceding claim, wherein the second one or more graphical data comprises one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. 9. The method of any preceding claim, wherein the sampling resolution of the first plurality of graphical data comprises the minimum resolution associated with the first plurality of graphical data. DB1/ 147040794.2 118 Attorney Ref. No.: 115834-5037-WO 10. The method of any preceding claim, wherein the first modality comprises a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof. 11. The method of any preceding claim, wherein the first codec comprises a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof. 12. The method of any preceding claim, wherein the first codec is selected by the first computer system. 13. The method of any preceding claim, wherein the plurality of sequence scans comprises between 2 scans and 100 scans. 14. The method according to any one of claims 1-12, wherein the plurality of sequence scans comprises N scans, and wherein:
Figure imgf000121_0001
M is a number of digital assets associated with the second one or more graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. 15. The method according to any one of claims 1-12, wherein the plurality of sequence scans comprises N scans, and wherein:
Figure imgf000121_0002
M is a number of digital assets associated with the second one or more graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. DB1/ 147040794.2 119 Attorney Ref. No.: 115834-5037-WO 16. The method of any preceding claim, wherein the first plurality of characteristics further comprises one or more ground truth labels with respect to the first modality for the first plurality of graphical data. 17. The method of claim 16, wherein the using B) further comprising training the feature extraction model against the one or more ground truth labels. 18. The method of any preceding claim, wherein the second plurality of characteristics comprises a sampling resolution of the second one or more graphical data, and wherein the plurality of sequence scans comprises N scans based, at least in part, on the sampling resolution of the second one or more graphical data. 19. The method of any preceding claim, wherein the second plurality of characteristics further comprises a respective capacity of the communication network. 20. The method of any preceding claim, wherein the second plurality of characteristics further comprises a respective capacity of the second computer system. 21. The method of any preceding claim, wherein the feature extraction model comprises a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof. 22. The method of any preceding claim, wherein the feature extraction model comprises a neutral network model, a support machine mode, a Naïve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph-based model, or a combination thereof. 23. The method of any preceding claim, wherein the first sequence scan comprises less than 5% of the encoded byte stream. 24. The method of any preceding claim, wherein the first sequence scan comprises less than 5% of a final terminal scan. 25. The method of any preceding claim, wherein the desired threshold performance comprises a threshold inter-image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation. DB1/ 147040794.2 120 Attorney Ref. No.: 115834-5037-WO 26. The method of any preceding claim, wherein: the encoded byte stream is associated with a first file size of the final terminal scan in the plurality of sequence scans, and a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan. 27. A computer system, the computer system comprising one or more processors, a controller, wherein at least one program is non-transiently stored in the controller and executable by the controller, the at least one program causing the controller to perform a method comprising: A) obtaining, in electronic form, a training data set comprising a first plurality of graphical data, wherein the first plurality of graphical data is defined by a first plurality of characteristics comprising: a sampling resolution of the first plurality of graphical data, and a first modality associated with a capture of the first plurality of graphical data; B) using the first plurality of graphical data to train a feature extraction model, thereby obtaining (i) a trained feature extraction model and (ii) performance data, the performance data comprising a plurality of training resolutions, wherein each respective training resolution in the plurality of training resolutions is less than the sampling resolution and is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances; C) receiving a request to evaluate a second one or more graphical data at a second computer system, wherein the second one or more graphical data is defined by a second plurality of characteristics comprising: the first modality associated with a capture of the second one or more graphical data, and a desired threshold performance; D) encoding, in accordance with a corresponding encoder of a first codec, the one or more graphical data, thereby forming an encoded byte stream comprising a plurality of sequence scans through the sampling resolution, wherein: each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans, and DB1/ 147040794.2 121 Attorney Ref. No.: 115834-5037-WO each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans; E) matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data; and F) communicating, via a communication network, in accordance with the first codec, to a second computer system, the encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream at the second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, thereby optimizing decoding of graphical data. 28. A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computer system, cause the computer system to perform a method comprising: A) obtaining, in electronic form, a training data set comprising a first plurality of graphical data, wherein the first plurality of graphical data is defined by a first plurality of characteristics comprising: a sampling resolution of the first plurality of graphical data, and a first modality associated with a capture of the first plurality of graphical data; B) using the first plurality of graphical data to train a feature extraction model, thereby obtaining (i) a trained feature extraction model and (ii) performance data, the performance data comprising a plurality of training resolutions, wherein each respective training resolution in the plurality of training resolutions is less than the sampling resolution and is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances; C) receiving a request to evaluate a second one or more graphical data at a second computer system, wherein the second one or more graphical data is defined by a second plurality of characteristics comprising: the first modality associated with a capture of the second one or more graphical data, and a desired threshold performance; DB1/ 147040794.2 122 Attorney Ref. No.: 115834-5037-WO D) encoding, in accordance with a corresponding encoder of a first codec, the one or more graphical data, thereby forming an encoded byte stream comprising a plurality of sequence scans through the sampling resolution, wherein: each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans, and each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans; E) matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data; and F) communicating, via a communication network, in accordance with the first codec, to a second computer system, the encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream at the second computer system when the decoding decodes a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, thereby optimizing decoding of graphical data. 29. A method for optimizing encoding of graphical data, the method comprising: at a first computer system comprising a processor and memory: A) obtaining, in electronic form, a training data set comprising a first plurality of graphical data, wherein the first plurality of graphical data is defined by a first plurality of characteristics comprising: a sampling resolution of the first plurality of graphical data, and a first modality associated with a capture of the first plurality of graphical data; B) using the first plurality of graphical data to train a feature extraction model, thereby obtaining (i) a trained feature extraction model and (ii) performance data, the performance data comprising a plurality of training resolutions, wherein each respective training resolution in the plurality of training resolutions is less than the sampling resolution and is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances; DB1/ 147040794.2 123 Attorney Ref. No.: 115834-5037-WO C) receiving a request to evaluate a second one or more graphical data at a second computer system, wherein the second one or more graphical data is defined by a second plurality of characteristics comprising: the first modality associated with a capture of the second one or more graphical data, and a desired threshold performance; and D) encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data, thereby forming an encoded byte stream comprising a plurality of sequence scans through the sampling resolution, wherein: each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans, and each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans, thereby optimizing encoding of graphical data. 30. The method of claim 29, wherein the first plurality of graphical data comprises one or more images taken of a first biological sample. 31. The method of claim 30, wherein the second one or more graphical data comprises one or more images taken of a second biological sample different than the first biological sample. 32. The method according to any one of claims 29-31, wherein the first plurality of graphical data comprises two-dimensional (2D) graphical data, three-dimensional (3D) graphical data, four-dimensional (4D) data, or a combination thereof. 33. The method according to any one of claims 29-31, wherein the second one or more graphical data comprises 3D graphical data and each scan in the plurality of sequence scans of the 3D graphical data comprises a respective 2D layer of the 3D graphical data. 34. The method according to any one of claims 29-33, wherein the first plurality of graphical data comprises one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. DB1/ 147040794.2 124 Attorney Ref. No.: 115834-5037-WO 35. The method of claim 34, wherein the first plurality of graphical data comprises 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images. 36. The method according to any one of claims 29-35, wherein the second one or more graphical data comprises one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. 37. The method according to any one of claims 29-36, wherein the sampling resolution of the first plurality of graphical data comprises the minimum resolution associated with the first plurality of graphical data. 38. The method according to any one of claims 29-37, wherein the first modality comprises a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof. 39. The method according to any one of claims 29-38, wherein the first codec comprises a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof. 40. The method according to any one of claims 29-39, wherein the first codec is selected by the first computer system. 41. The method according to any one of claims 29-40, wherein the plurality of sequence scans comprises between 2 scans and 100 scans. 42. The method according to any one of claims 29-40, wherein the plurality of sequence scans comprises N scans, and wherein:
Figure imgf000127_0001
M is a number of digital assets associated with the second one or more graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and DB1/ 147040794.2 125 Attorney Ref. No.: 115834-5037-WO ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. 43. The method according to any one of claims 29-40, wherein the plurality of sequence scans comprises N scans, and wherein:
Figure imgf000128_0001
M is a number of digital assets associated with the second one or more graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. 44. The method according to any one of claims 29-43, wherein the first plurality of characteristics further comprises one or more ground truth labels with respect to the first modality for the first plurality of graphical data. 45. The method of claim 44, wherein the using B) further comprising training the feature extraction model against the one or more ground truth labels. 46. The method according to any one of claims 29-45, wherein the second plurality of characteristics comprises a sampling resolution of the second one or more graphical data, and wherein the plurality of sequence scans comprises N scans based, at least in part, on the sampling resolution of the second one or more graphical data. 47. The method according to any one of claims 29-46, wherein the second plurality of characteristics further comprises a respective capacity of the communication network. 48. The method according to any one of claims 29-47, wherein the second plurality of characteristics further comprises a respective capacity of the second computer system. 49. The method according to any one of claims 29-48, wherein the feature extraction model comprises a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof. 50. The method according to any one of claims 29-49, wherein the feature extraction model comprises a neutral network model, a support machine mode, a Naïve Bayes model, a nearest DB1/ 147040794.2 126 Attorney Ref. No.: 115834-5037-WO neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph- based model, or a combination thereof. 51. The method according to any one of claims 29-50, wherein the desired threshold performance comprises a threshold inter-image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation. 52. The method according to any one of claims 29-51, wherein: the encoded byte stream is associated with a first file size of the final terminal scan in the plurality of sequence scans, and a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan. 53. A computer system, the computer system comprising one or more processors, a controller, wherein at least one program is non-transiently stored in the controller and executable by the controller, the at least one program causing the controller to perform a method comprising: A) obtaining, in electronic form, a training data set comprising a first plurality of graphical data, wherein the first plurality of graphical data is defined by a first plurality of characteristics comprising: a sampling resolution of the first plurality of graphical data, and a first modality associated with a capture of the first plurality of graphical data; B) using the first plurality of graphical data to train a feature extraction model, thereby obtaining (i) a trained feature extraction model and (ii) performance data, the performance data comprising a plurality of training resolutions, wherein each respective training resolution in the plurality of training resolutions is less than the sampling resolution and is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances; C) receiving a request to evaluate a second one or more graphical data at a second computer system, wherein the second one or more graphical data is defined by a second plurality of characteristics comprising: the first modality associated with a capture of the second one or more graphical data, and DB1/ 147040794.2 127 Attorney Ref. No.: 115834-5037-WO a desired threshold performance; and D) encoding, in accordance with a corresponding encoder of a first codec, the one or more graphical data, thereby forming an encoded byte stream comprising a plurality of sequence scans through the sampling resolution, wherein: each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans, and each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans, thereby optimizing encoding of graphical data. 54. A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computer system, cause the computer system to perform a method comprising: A) obtaining, in electronic form, a training data set comprising a first plurality of graphical data, wherein the first plurality of graphical data is defined by a first plurality of characteristics comprising: a sampling resolution of the first plurality of graphical data, and a first modality associated with a capture of the first plurality of graphical data; B) using the first plurality of graphical data to train a feature extraction, thereby obtaining (i) a trained feature extraction model and (ii) performance data, the performance data comprising a plurality of training resolutions, wherein each respective training resolution in the plurality of training resolutions is less than the sampling resolution and is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances; C) receiving a request to evaluate a second one or more graphical data at a second computer system, wherein the second one or more graphical data is defined by a second plurality of characteristics comprising: the first modality associated with a capture of the second one or more graphical data, and a desired threshold performance; and DB1/ 147040794.2 128 Attorney Ref. No.: 115834-5037-WO D) encoding, in accordance with a corresponding encoder of a first codec, the one or more graphical data, thereby forming an encoded byte stream comprising a plurality of sequence scans through the sampling resolution, wherein: each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans, and each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans, thereby optimizing encoding of graphical data. 55. A method for optimizing communication of graphical data, the method comprising: at a first computer system comprising a processor and memory: A) obtaining, in electronic form, a training data set comprising a first plurality of graphical data, wherein the first plurality of graphical data is defined by a first plurality of characteristics comprising: a sampling resolution of the first plurality of graphical data, and a first modality associated with a capture of the first plurality of graphical data; B) using the first plurality of graphical data to train a feature extraction model, thereby obtaining (i) a trained feature extraction model and (ii) performance data, the performance data comprising a plurality of training resolutions, wherein each respective training resolution in the plurality of training resolutions is less than the sampling resolution and is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances; C) receiving a request to evaluate a second one or more graphical data at a second computer system, wherein the second one or more graphical data is defined by a second plurality of characteristics comprising: the first modality associated with a capture of the second one or more graphical data, and a desired threshold performance; DB1/ 147040794.2 129 Attorney Ref. No.: 115834-5037-WO D) encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data, thereby forming an encoded byte stream comprising a plurality of sequence scans through the sampling resolution, wherein: each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans, and each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans; E) matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data; F) communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream; and G) terminating, in accordance with a determination that decoding of the encoded byte stream at the second computer system satisfies a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, the communicating F), thereby optimizing communication of graphical data. 56. The method of claim 55, wherein the first plurality of graphical data comprises one or more images taken of a first biological sample. 57. The method of claim 56, wherein the second one or more graphical data comprises one or more images taken of a second biological sample different than the first biological sample. 58. The method according to any one of claims 55-57, wherein the first plurality of graphical data comprises two-dimensional (2D) graphical data, three-dimensional (3D) graphical data, four-dimensional (4D) data, or a combination thereof. 59. The method according to any one of claims 55-57, wherein the second one or more graphical data comprises 3D graphical data and each scan in the plurality of sequence scans of the 3D graphical data comprises a respective 2D layer of the 3D graphical data. 60. The method according to any one of claims 55-59, wherein the first plurality of graphical data comprises one or more digital images, one or more digital videos, one or more 2D maps, DB1/ 147040794.2 130 Attorney Ref. No.: 115834-5037-WO one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. 61. The method of claim 60, wherein the first plurality of graphical data comprises 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images. 62. The method according to any one of claims 55-61, wherein the second one or more graphical data comprises one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. 63. The method according to any one of claims 55-62, wherein the sampling resolution of the first plurality of graphical data comprises the minimum resolution associated with the first plurality of graphical data. 64. The method according to any one of claims 55-63, wherein the first modality comprises a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof. 65. The method according to any one of claims 55-64, wherein the first codec comprises a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof. 66. The method according to any one of claims 55-65, wherein the first codec is selected by the first computer system. 67. The method according to any one of claims 55-66, wherein the plurality of sequence scans comprises between 2 scans and 100 scans. 68. The method according to any one of claims 55-66, wherein the plurality of sequence scans comprises N scans, and wherein:
Figure imgf000133_0001
M is a number of digital assets associated with the second one or more graphical data, DB1/ 147040794.2 131 Attorney Ref. No.: 115834-5037-WO ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. 69. The method according to any one of claims 55-66, wherein the plurality of sequence scans comprises N scans, and wherein:
Figure imgf000134_0001
M is a number of digital assets associated with the second one or more graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and ^^^^ ^^^^ ^^^^ ^^^^ is the minimum resolution of the initial terminal scan in the plurality of sequenced scans. 70. The method according to any one of claims 55-69, wherein the first plurality of characteristics further comprises one or more ground truth labels with respect to the first modality for the first plurality of graphical data. 71. The method of claim 70, wherein the using B) further comprising training the feature extraction model against the one or more ground truth labels. 72. The method according to any one of claims 55-71, wherein the second plurality of characteristics comprises a sampling resolution of the second one or more graphical data, and wherein the plurality of sequence scans comprises N scans based, at least in part, on the sampling resolution of the second one or more graphical data. 73. The method according to any one of claims 55-72, wherein the second plurality of characteristics further comprises a respective capacity of the communication network. 74. The method according to any one of claims 55-73, wherein the second plurality of characteristics further comprises a respective capacity of the second computer system. 75. The method according to any one of claims 55-74, wherein the feature extraction model comprises a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof. DB1/ 147040794.2 132 Attorney Ref. No.: 115834-5037-WO 76. The method according to any one of claims 55-75, wherein the feature extraction model comprises a neutral network model, a support machine mode, a Naïve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph- based model, or a combination thereof. 77. The method according to any one of claims 55-76, wherein the first sequence scan comprises less than 5% of the encoded byte stream. 78. The method according to any one of claims 55-77, wherein the first sequence scan comprises less than 5% of a final terminal scan. 79. The method according to any one of claims 55-78, wherein the desired threshold performance comprises a threshold inter-image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation. 80. The method according to any one of claims 55-79, wherein: the encoded byte stream is associated with a first file size of the final terminal scan in the plurality of sequence scans, and a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan. 81. A computer system, the computer system comprising one or more processors, a controller, wherein at least one program is non-transiently stored in the controller and executable by the controller, the at least one program causing the controller to perform a method comprising: A) obtaining, in electronic form, a training data set comprising a first plurality of graphical data, wherein the first plurality of graphical data is defined by a first plurality of characteristics comprising: a sampling resolution of the first plurality of graphical data, a first modality associated with a capture of the first plurality of graphical data; B) using the first plurality of graphical data to train a feature extraction model, thereby obtaining (i) a trained feature extraction model and (ii) performance data, the performance data comprising a plurality of training resolutions, wherein each respective training resolution in the plurality of training resolutions is less than the sampling resolution DB1/ 147040794.2 133 Attorney Ref. No.: 115834-5037-WO and is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances; C) receiving a request to evaluate a second one or more graphical data at a second computer system, wherein the second one or more graphical data is defined by a second plurality of characteristics comprising: the first modality associated with a capture of the second one or more graphical data, and a desired threshold performance; D) encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data, thereby forming an encoded byte stream comprising a plurality of sequence scans through the sampling resolution, wherein: each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans, and each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans; and E) matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data; and F) communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream; and G) terminating, in accordance with a determination that decoding of the encoded byte stream at the second computer system satisfies a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, the communicating F), thereby optimizing communication of graphical data. 82. A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computer system, cause the computer system to perform a method comprising: A) obtaining, in electronic form, a training data set comprising a first plurality of graphical data, wherein the first plurality of graphical data is defined by a first plurality of characteristics comprising: a sampling resolution of the first plurality of graphical data, and DB1/ 147040794.2 134 Attorney Ref. No.: 115834-5037-WO a first modality associated with a capture of the first plurality of graphical data; B) using the first plurality of graphical data to train a feature extraction model, thereby obtaining (i) a trained feature extraction model and (ii) performance data, the performance data comprising a plurality of training resolutions, wherein each respective training resolution in the plurality of training resolutions is less than the sampling resolution and is associated with a corresponding threshold performance for the trained feature extraction model in a plurality of threshold performances; C) receiving a request to evaluate a second one or more graphical data at a second computer system, wherein the second one or more graphical data is defined by a second plurality of characteristics comprising: the first modality associated with a capture of the second one or more graphical data, and a desired threshold performance; D) encoding, in accordance with a corresponding encoder of a first codec, the second one or more graphical data, thereby forming an encoded byte stream comprising a plurality of sequence scans through the sampling resolution, wherein: each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans, and each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans; and E) matching the desired threshold performance to a first training resolution in the plurality of training resolutions in the performance data; and F) communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream; and G) terminating, in accordance with a determination that decoding of the encoded byte stream at the second computer system satisfies a first sequence scan in the plurality of sequence scans that matches or exceeds the first training resolution, the communicating F), thereby optimizing communication of graphical data. DB1/ 147040794.2 135 Attorney Ref. No.: 115834-5037-WO 83. A method for optimizing decoding of graphical data, the method comprising: at a first computer system comprising a processor and memory: A) obtaining, in electronic form, a first one or more graphical data, wherein the first one or more graphical data is defined by a first plurality of characteristics comprising: a sampling resolution of the first one or more of graphical data, and a first modality associated with a capture of the first one or more of graphical data; B) forming a byte stream map for the first one or more graphical data, the byte stream map comprising a plurality of sequence scans through the sampling resolution, wherein: each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans, and each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans; C) receiving, from a second computer system, a request to evaluate the first one or more graphical data at a first resolution, wherein the first resolution is selected in accordance with a desired threshold performance from performance data, the performance data comprising a plurality of training resolutions, wherein each respective training resolution in the plurality of training resolutions is less than the sampling resolution and is associated with a corresponding threshold performance for a trained feature extraction model in a plurality of threshold performances, each threshold performance in the plurality of threshold performances associated with the first modality associated with a capture of a second plurality of graphical data; D) matching the first resolution to a first sequence scan in the plurality of sequence scans in accordance with the byte stream map; E) encoding, in accordance with a corresponding encoder of a first codec, the first one or more graphical data, thereby forming an encoded byte stream comprising the plurality of sequence scans through at least the first resolution; and F) communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream at the second computer system when the DB1/ 147040794.2 136 Attorney Ref. No.: 115834-5037-WO decoding decodes the first sequence scan in the plurality of sequence scans, thereby optimizing decoding of graphical data. 84. The method of claim 83, wherein the first one or more of graphical data comprises one or more images taken of a first biological sample. 85. The method of claim 84, wherein the first one or more graphical data comprises one or more images taken of a second biological sample different than the first biological sample. 86. The method of any one of claims 83-85, wherein the second plurality of graphical data comprises two-dimensional (2D) graphical data, three-dimensional (3D) graphical data, four- dimensional (4D) data, or a combination thereof. 87. The method according to any one of claims 83-86, wherein the first one or more graphical data comprises 3D graphical data and each scan in the plurality of sequence scans of the 3D graphical data comprises a respective 2D layer of the 3D graphical data. 88. The method according to any one of claims 83-87, wherein the second plurality of graphical data comprises one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. 89. The method of claim 88, wherein the second plurality of graphical data comprises 50 or more digital images, 100 or more digital images, 1,000 or more digital images, 10,000 or more digital images, or 100,000 or more digital images. 90. The method according to any one of claims 83-89, wherein the first one or more graphical data comprises one or more digital images, one or more digital videos, one or more 2D maps, one or more 3D maps, one or more dense point clouds, one or more textured meshes, one or more cryptographic non-fungible token assets, or a combination thereof. 91. The method according to any one of claims 83-90, wherein the sampling resolution of the first one or more graphical data comprises the minimum resolution associated with the first one or more graphical data. 92. The method according to any one of claims 83-91, wherein the first modality comprises a computer tomography (CT) modality, a digital pathology modality, a magnetic resonance DB1/ 147040794.2 137 Attorney Ref. No.: 115834-5037-WO imaging (MRI) modality, a positron emission tomography (PET) modality, a radiograph modality, a single-photon emission (SPE) modality, a sonography modality, or a combination thereof. 93. The method according to any one of claims 83-92, wherein the first codec comprises a predictive codec, an embedded codec, a sub-band codec, a block-based codec, a layered codec, a lossless codec, a lossy codec, or a combination thereof. 94. The method according to any one of claims 83-93, wherein the first codec is selected by the second computer system. 95. The method according to any one of claims 83-94, wherein the plurality of sequence scans comprises between 2 scans and 100 scans. 96. The method according to any one of claims 83-95, wherein the plurality of sequence scans comprises N scans, and wherein: ^^^^ is a number of digital assets associated with the second one or
Figure imgf000140_0001
more graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and ^^^^ ^^^^ ^^^^ ^^^^ the minimum resolution of the initial terminal scan in the plurality of sequenced scans. 97. The method according to any one of claims 83-95, wherein the plurality of sequence scans comprises N scans, and wherein: ^^^^ = − 1is a number of digital assets associated with the second one or
Figure imgf000140_0002
more graphical data, ^^^^ ^^^^ ∀ ^^^^ ∈ {1,2, .. , ^^^^} is a native resolution associated with the second one or more graphical data, and ^^^^ ^^^^ ^^^^ ^^^^ the minimum resolution of the initial terminal scan in the plurality of sequenced scans. 98. The method according to any one of claims 83-97, wherein the first plurality of characteristics further comprises one or more ground truth labels with respect to the first modality for the first one or more graphical data. DB1/ 147040794.2 138 Attorney Ref. No.: 115834-5037-WO 99. The method of claim 98, wherein, prior to the receiving C), the method further comprises training the feature extraction model against the one or more ground truth labels. 100. The method according to any one of claims 83-99, wherein the first plurality of characteristics comprises a sampling resolution of the first one or more graphical data, and wherein the plurality of sequence scans comprises N scans based, at least in part, on the sampling resolution of the first one or more graphical data. 101. The method according to any one of claims 83-100, wherein the first plurality of characteristics further comprises a respective capacity of the communication network. 102. The method according to any one of claims 83-101, wherein the first plurality of characteristics further comprises a respective capacity of the second computer system. 103. The method according to any one of claims 83-102, wherein the feature extraction model comprises a segmentation model, a classification model, a regression model, a statistical model, or a combination thereof. 104. The method according to any one of claims 83-103, wherein the feature extraction model comprises a neutral network model, a support machine mode, a Naïve Bayes model, a nearest neighbor model, a boosted trees model, a random forest model, a decision tree model, a clustering model, an extreme gradient boost (XGBoost) model, a convolutional or graph- based model, or a combination thereof. 105. The method according to any one of claims 83-104, wherein the first sequence scan comprises less than 5% of the encoded byte stream. 106. The method according to any one of claims 83-105, wherein the first sequence scan comprises less than 5% of a final terminal scan. 107. The method according to any one of claims 83-106, wherein the desired threshold performance comprises a threshold inter-image correlation, a threshold intra-image correlation, or a threshold intra-and-inter-image correlation. 108. The method according to any one of claims 83-107, wherein: the encoded byte stream is associated with a first file size of the final terminal scan in the plurality of sequence scans, and DB1/ 147040794.2 139 Attorney Ref. No.: 115834-5037-WO a second file size of the first sequence scan in the plurality of sequence scans is less than the first file size of the final terminal scan. 109. A non-transitory computer readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by a computer system, cause the computer system to perform a method comprising: A) obtaining, in electronic form, a first one or more graphical data, wherein the first one or more graphical data is defined by a first plurality of characteristics comprising: a sampling resolution of the first one or more of graphical data, and a first modality associated with a capture of the first one or more of graphical data; B) forming a byte stream map for the first one or more graphical data, the byte stream map comprising a plurality of sequence scans through the sampling resolution, wherein: each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans, and each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans; C) receiving, from a second computer system, a request to evaluate the first one or more graphical data at a first resolution, wherein the first resolution is selected in accordance with a desired threshold performance from performance data, the performance data comprising a plurality of training resolutions, wherein each respective training resolution in the plurality of training resolutions is less than the sampling resolution and is associated with a corresponding threshold performance for a trained feature extraction model in a plurality of threshold performances, each threshold performance in the plurality of threshold performances associated with the first modality associated with a capture of a second plurality of graphical data; D) matching the first resolution to a first sequence scan in the plurality of sequence scans in accordance with the byte stream map; E) encoding, in accordance with a corresponding encoder of a first codec, the first one or more graphical data, thereby forming an encoded byte stream comprising the plurality of sequence scans through at least the first resolution; and DB1/ 147040794.2 140 Attorney Ref. No.: 115834-5037-WO F) communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream at the second computer system when the decoding decodes the first sequence scan in the plurality of sequence scans, thereby optimizing decoding of graphical data. 110. A method for optimizing encoding of graphical data, the method comprising: at a first computer system comprising a processor and memory: A) obtaining, in electronic form, a first one or more graphical data, wherein the first one or more graphical data is defined by a first plurality of characteristics comprising: a sampling resolution of the first one or more of graphical data, and a first modality associated with a capture of the first one or more of graphical data; B) forming a byte stream map for the first one or more graphical data, the byte stream map comprising a plurality of sequence scans through the sampling resolution, wherein: each respective sequence scan in the plurality of sequence scans is associated with a unique rank in a rank order that defines a sequence of respective sequence scans in the plurality of sequence scans, and each sequence scan in the sequence of the plurality of sequence scans has a corresponding progressive resolution that is based, at least in part, on a resolution of a preceding scan in the plurality of sequence scans or a predetermined resolution associated with an initial terminal scan in the plurality of sequence scans; C) receiving, from a second computer system, a request to evaluate the first one or more graphical data at a first resolution, wherein the first resolution is selected in accordance with a desired threshold performance from performance data, the performance data comprising a plurality of training resolutions, wherein each respective training resolution in the plurality of training resolutions is less than the sampling resolution and is associated with a corresponding threshold performance for a trained feature extraction model in a plurality of threshold performances, each threshold performance in the plurality of threshold performances associated with the first modality associated with a capture of a second plurality of graphical data; D) matching the first resolution to a first sequence scan in the plurality of sequence scans in accordance with the byte stream map; DB1/ 147040794.2 141 Attorney Ref. No.: 115834-5037-WO E) encoding, in accordance with a corresponding encoder of a first codec, the first one or more graphical data, thereby forming an encoded byte stream comprising the plurality of sequence scans through at least the first resolution; and F) communicating, via a communication network, in accordance with the first codec, to the second computer system, the encoded byte stream and one or more instructions to terminate decoding of the encoded byte stream at the second computer system when the decoding decodes the first sequence scan in the plurality of sequence scans, thereby optimizing decoding of graphical data. DB1/ 147040794.2 142
PCT/US2024/028956 2023-05-11 2024-05-10 Systems and methods for high-throughput analysis for graphical data Pending WO2024233969A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US202363501552P 2023-05-11 2023-05-11
US63/501,552 2023-05-11

Publications (1)

Publication Number Publication Date
WO2024233969A1 true WO2024233969A1 (en) 2024-11-14

Family

ID=93431215

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2024/028956 Pending WO2024233969A1 (en) 2023-05-11 2024-05-10 Systems and methods for high-throughput analysis for graphical data

Country Status (1)

Country Link
WO (1) WO2024233969A1 (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN120163976A (en) * 2025-02-26 2025-06-17 中南大学湘雅医院 A method and system for image recognition of urinary stones

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190385288A1 (en) * 2015-09-17 2019-12-19 Michael Edwin Stewart Methods and Apparatus for Enhancing Optical Images and Parametric Databases
US20220215918A1 (en) * 2018-11-21 2022-07-07 Enlitic, Inc. Re-training a model for abnormality detection in medical scans based on a re-contrasted training set
US20220260826A1 (en) * 2018-12-18 2022-08-18 Pathware Inc. Computational microscopy based-system and method for automated imaging and analysis of pathology specimens
WO2022265883A1 (en) * 2021-06-17 2022-12-22 Ventana Medical Systems, Inc. Digital synthesis of histological stains using multiplexed immunofluorescence imaging

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190385288A1 (en) * 2015-09-17 2019-12-19 Michael Edwin Stewart Methods and Apparatus for Enhancing Optical Images and Parametric Databases
US20220215918A1 (en) * 2018-11-21 2022-07-07 Enlitic, Inc. Re-training a model for abnormality detection in medical scans based on a re-contrasted training set
US20220260826A1 (en) * 2018-12-18 2022-08-18 Pathware Inc. Computational microscopy based-system and method for automated imaging and analysis of pathology specimens
WO2022265883A1 (en) * 2021-06-17 2022-12-22 Ventana Medical Systems, Inc. Digital synthesis of histological stains using multiplexed immunofluorescence imaging

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN120163976A (en) * 2025-02-26 2025-06-17 中南大学湘雅医院 A method and system for image recognition of urinary stones

Similar Documents

Publication Publication Date Title
CN113763314B (en) Systems and methods for image segmentation and classification using deep reduced convolutional neural networks
AlGhamdi et al. DU-Net: Convolutional network for the detection of arterial calcifications in mammograms
KR101857624B1 (en) Medical diagnosis method applied clinical information and apparatus using the same
US8811697B2 (en) Data transmission in remote computer assisted detection
US20200320685A1 (en) Automated classification and taxonomy of 3d teeth data using deep learning methods
CN109978037A (en) Image processing method, model training method, device and storage medium
Anas et al. Advancing breast cancer detection: enhancing YOLOv5 network for accurate classification in mammogram images
WO2016057960A1 (en) Apparatus, system and method for cloud based diagnostics and image archiving and retrieval
Sadad et al. Internet of medical things embedding deep learning with data augmentation for mammogram density classification
Zheng et al. Deep learning based automatic segmentation of pathological kidney in CT: local versus global image context
Ikechukwu et al. CX-Net: an efficient ensemble semantic deep neural network for ROI identification from chest-x-ray images for COPD diagnosis
US20180365876A1 (en) Method, apparatus and system for spine labeling
Malekmohammadi et al. Mass segmentation in automated breast ultrasound using an enhanced attentive UNet
US11455755B2 (en) Methods and apparatus for neural network based image reconstruction
Cai et al. Identifying architectural distortion in mammogram images via a se-densenet model and twice transfer learning
Valizadeh et al. The Progress of Medical Image Semantic Segmentation Methods for Application in COVID‐19 Detection
Khaledyan et al. WATUNet: a deep neural network for segmentation of volumetric sweep imaging ultrasound
WO2024233969A1 (en) Systems and methods for high-throughput analysis for graphical data
WO2021197176A1 (en) Systems and methods for tumor characterization
CN118710897A (en) A method and system for automatic segmentation of clinical target area for brachytherapy of cervical cancer
Cifci et al. Deep learning algorithms for diagnosis of breast cancer with maximum likelihood estimation
Costa et al. Detection of architectural distortion with deep convolutional neural network and data augmentation of limited dataset
Tkachenko et al. A Mammography Data Management Application for Federated Learning
Ding et al. Deep transfer learning-based intelligent diagnosis of malignant tumors on mammography
Kumar Intracranial Hemorrhage Detection Using Deep Learning and Transfer Learning

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 24804367

Country of ref document: EP

Kind code of ref document: A1