[go: up one dir, main page]

WO2025024839A2 - Procédé et système d'enregistrement et de décodage de paramètres de l'activité cérébrale de chiens - Google Patents

Procédé et système d'enregistrement et de décodage de paramètres de l'activité cérébrale de chiens Download PDF

Info

Publication number
WO2025024839A2
WO2025024839A2 PCT/US2024/039970 US2024039970W WO2025024839A2 WO 2025024839 A2 WO2025024839 A2 WO 2025024839A2 US 2024039970 W US2024039970 W US 2024039970W WO 2025024839 A2 WO2025024839 A2 WO 2025024839A2
Authority
WO
WIPO (PCT)
Prior art keywords
canine
eeg
cap
bci
neurological
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/US2024/039970
Other languages
English (en)
Other versions
WO2025024839A3 (fr
Inventor
Akash KULGOD
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.)
Dognosis Inc
Original Assignee
Dognosis Inc
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 Dognosis Inc filed Critical Dognosis Inc
Publication of WO2025024839A2 publication Critical patent/WO2025024839A2/fr
Publication of WO2025024839A3 publication Critical patent/WO2025024839A3/fr
Pending legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
    • G06F3/011—Arrangements for interaction with the human body, e.g. for user immersion in virtual reality
    • G06F3/015—Input arrangements based on nervous system activity detection, e.g. brain waves [EEG] detection, electromyograms [EMG] detection, electrodermal response detection
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
    • G06F3/01—Input arrangements or combined input and output arrangements for interaction between user and computer
    • G06F3/011—Arrangements for interaction with the human body, e.g. for user immersion in virtual reality
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B2503/00—Evaluating a particular growth phase or type of persons or animals
    • A61B2503/40—Animals
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B2503/00—Evaluating a particular growth phase or type of persons or animals
    • A61B2503/42—Evaluating a particular growth phase or type of persons or animals for laboratory research
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
    • A61B5/165—Evaluating the state of mind, e.g. depression, anxiety
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/25—Bioelectric electrodes therefor
    • A61B5/251—Means for maintaining electrode contact with the body
    • A61B5/256—Wearable electrodes, e.g. having straps or bands
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
    • A61B5/25—Bioelectric electrodes therefor
    • A61B5/279—Bioelectric electrodes therefor specially adapted for particular uses
    • A61B5/291—Bioelectric electrodes therefor specially adapted for particular uses for electroencephalography [EEG]
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/40—Detecting, measuring or recording for evaluating the nervous system
    • A61B5/4076—Diagnosing or monitoring particular conditions of the nervous system
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
    • A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2203/00—Indexing scheme relating to G06F3/00 - G06F3/048
    • G06F2203/01—Indexing scheme relating to G06F3/01
    • G06F2203/011—Emotion or mood input determined on the basis of sensed human body parameters such as pulse, heart rate or beat, temperature of skin, facial expressions, iris, voice pitch, brain activity patterns

Definitions

  • the present invention relates to the field of electronic based measurement systems and analysis. More particularly, the present invention relates to a system for measuring brain signals of canines and analysis of the measurements thereof to detect predetermined canine behavioral states.
  • Canine cognitive neuroscience leverages tools such as functional Magnetic Resonance Imaging (fMRI) and Electroencephalography (EEG) to understand and predict canine mental activity and behavior.
  • fMRI Magnetic Resonance Imaging
  • EEG Electroencephalography
  • fMRI with its superior spatial resolution, allows for precise measurement of neural activity but suffers from poor temporal resolution, high costs and the need for immobility, limiting its practical use outside controlled laboratory settings.
  • EEG provides high temporal resolution and portability, making it suitable for real-life operational use despite challenges.
  • Form Factors Existing EEG systems are tailored for humans and do not have form factors suitable for the diverse head shapes and sizes of canine breeds.
  • Mobility Most current systems are not designed for free movement, often requiring bulky amplifiers and wired connections that restrict the dog's natural behavior.
  • BCI canine Brain-Computer Interface
  • BCI Brain-Computer Interface
  • the present invention has the unique ability to provide customized solutions to addresses the above-described challenges by providing a tailored cap uniquely suited for canine anatomy, which enables collection and reliable analysis of EEG and neuroimaging/other sensor relayed information for canine brain activity, even while the canine is in motion.
  • the system is adapted to ideally suit canine size and breed, tailored to closely follow canine anatomy, with a highly unique intelligent accompanying system to provide signal collection that is “de-noised”, while correlating the collected information to form predictive outputs related to canine neurologic and behavioral responses.
  • Working Dogs Enhancing the capabilities of detection dogs for explosives, narcotics, diseases, and other critical tasks. For instance, canine BCIs could improve selection processes by identifying neural patterns associated with key traits like learning ability and stress resilience. During training, real-time neural feedback could optimize training protocols. In deployment, BCIs could decode trained odor stimuli from brain signals, providing digital and multidimensional indications without the need for behavioral cues. • Research and Veterinary Use: Facilitating advanced studies in canine cognitive neuroscience and improving diagnostic and therapeutic practices in veterinary medicine. BCIs could provide insights into canine cognitive processes, emotional states, and health conditions, enabling more precise and personalized interventions.
  • the devices, systems, and/or methods of this invention provide for significant advancement in the field of canine cognitive neuroscience, promising to enhance our understanding, ultimately, of canine neurologic and behavioral responses and improving human-canine interactions across various domains.
  • the present invention relates to a system and method for obtaining canine brain signals, analyzing same and determining a predefined behavioral state of the canine based on said analysis.
  • This invention therefore provides a device, system and/or method for determining a neurological or behavioral state of a canine.
  • this invention provides a customized brain-computer interface (BCI) cap for canines, the cap comprising:
  • a transmitter operationally connected to the analog to digital converter, which is optionally wireless; and a power source for and operationally connected to the EEG sensors, the ampli bomb, the analog to digital converter and the wireless transmitter; wherein the canine scalp and the fitted cap is attached to maintain consistent contact therewith to promote reliable EEG signal collection from the plurality of EEG sensors.
  • the invention relates to a customized brain-computer interface (BCI) cap for canines.
  • BCI brain-computer interface
  • the term “cap” is meant to convey a structure that substantially covers the majority of the canine skull, or in some embodiments, sufficiently covers landmark regions of the skull that houses the canine brain, at least covering its most apical region of the skull for same.
  • the “cap” may also be referred to as a helmet or head gear, or skull covering, and the like, and is uniquely specifically designed to optimally bring the sensors described herein into contact with the scalp region of the canine, proximal to the area of the skull housing the brain, and in some aspects, specific anatomic regions of placement of the sensors are provided that have been found to provide reliability in EEG sensor collected data and its analysis.
  • fitted cap as described herein in reference to the BCI caps of the invention will in some aspects relate to the ability to maintain the desired contact between the indicated sensors and the scalp of the canine in an indicated region, to promote the contact therebetween.
  • fitted caps may be prepared as further described herein, taking into consideration the various means to best provided a “fitted” gear, for example, referring to Figures 1-8 in this regard and accompanying description of same provided herein.
  • the customized cap uniquely provides for acquiring good signal quality data, as well as large amounts of data, that was heretofore unattainable.
  • EEG sensors refer to sensors (wet or dry or combination thereof) capable of collecting EEG signals when applied on a subject, (e.g., applied on the head of a subject, typically collecting signals from the brain of the subject) and refers to the myriad of commercially available and published sensors, and it is noted that the term EEG sensor is an art recognized term.
  • amplifier refers to an electronic device capable of increasing the voltage, current, or power of a signal and will encompass any commercially available or published device, and it is noted that the term is an art recognized term.
  • transmitter refers to an electronic component capable of transmitting signals and will encompass any commercially available or published device, and it is noted that the term is an art recognized term.
  • wireless refers to transmitting signals over invisible (e.g., radio) waves instead of wires and will encompass any commercially available or published device, and it is noted that the term is an art recognized term.
  • the cap, systems and any of the devices described herein as being enabled for wireless transmission may comprise communications electronics, which transmit the data to an external device for storage and/or analysis and will also be appreciated to refer to the art recognized term.
  • power source may refer to any applicable source of electrical energy, such as a battery, or power cable to connect to traditional power sources, and will encompass any commercially available or published device, and it is noted that the term is an art recognized term.
  • signal collection refers to collecting, measuring and/or analyzing signals made accessible by means of appropriate sensors and will encompass any commercially available or published device, and it is noted that the term is an art recognized term.
  • the cap is comprised of a polymer resin, which in some embodiments, is comprised of thermoplastic polyurethane, polylactic acid or combinations thereof. In some embodiments, any appropriate material that enables good fit achievement, promoting sensor stabilization and contact with the indicated scalp regions of the canine is envisioned.
  • the specialized compartments comprise grooves or indentations sized to accommodate insertion of said plurality of EEG sensors therewithin in a fitted manner.
  • compartment refers to holes, bores, grooves, slots, dimples, etc., or any containment structure, capable of housing the sensors therewithin, while enabling their functioning to occur.
  • the cap substantially covers the skull region housing the brain of the canine or in some embodiments, substantially covers landmark regions of the brain. In some embodiments, the cap substantially covers the majority of the skull region housing the brain of the canine. [00030] In other embodiments, the cap further comprises straps promoting positioning of said cap securely so that consistent contact between said canine scalp and said fitted cap is maintained.
  • the term "substanti covers the skull region housing the brain” refers to (describe using apical portion of the skull region, in terms of from about 60-85%, covering brain regions from which EEG signals optimally collected).
  • straps refers to attachment members capable of attaching the cap to the head of the canine and will encompass any commercially available strap or belt or the like, and it is noted that the term is an art recognized term.
  • the specialized compartments accommodate containment of two or three midline located EEG electrodes to be placed proximally to the sagittal crest of the canine skull, wherein said midline located EEG electrodes comprise Fz Cz and Pz electrodes, and in some embodiments, the midline located Fz electrode is proximally placed to a landmark glabellar region in said canine skull and in some embodiments, the midline located Pz electrode is proximally placed to a landmark sagittal crest or landmark external occipital protuberance of said canine skull.
  • the midline located Cz electrode is placed midway in location between said Fz electrode and said Cz electrode.
  • two of said specialized compartments each accommodate containment of EEG electrodes, which together form an equilateral triangle in position between said Fz electrode and said Cz electrode. In other embodiments, two of said specialized compartments each accommodate containment of EEG electrodes which together form an equilateral triangle in position between said Cz electrode and said Pz electrode. In other embodiments, two of said specialized compartments each accommodate containment of EEG electrodes Fpl and Fp2, which are located equidistant from and in the same plane as said Fz electrode. In other embodiments, two of said specialized compartments each accommodate containment of EEG electrodes to be placed proximally to the occipital region of the canine, optionally replacing EEG electrodes placed proximally to a sagittal crest.
  • EEG electrodes as referenced herein is to be understood to be synonymous and embodying every described embodied possibility with regard to an EEG sensor, as described herein.
  • the term “equidistant” refers to a substantially similar distance, or more or less equal distance between the two segments which are stated to be “equidistant” with respect thereto. It will be appreciated that a placement that is “equidistant” with reference points may however include positioning within 0.5 to about 8% closer or further away from one of the two reference points, while still being considered as satisfying the relative positioning qualification of being “equidistant”.
  • triangle refers to the positioning of the electrodes more or less on the corners of a polygon with three corners and three sides.
  • scaling of said cap accommodates and snugly fits different sized and breeds of dog, whereby said specialized compartments are positioned in locations to be consistent with said landmark glabellar region, saggital crest, external occipital protuberance and relative positioning of said plurality of electrodes.
  • electrodes positions may also accommodate sensing signals from other regions of the brain, including the prefrontal and other cortical regions, the temporal, parietal and occipital lobes, as well as deeper regions in the brain that may be accessed.
  • the electrodes may be positioned directly below the canine eye, to measure electrical activity from the canine olfactory bulb, which region may be more ideally located for sensing the activity from same, i.e. positioning the electrodes proximally to the lower orbicularis oculi region, which is to be considered an embodied aspect of the invention.
  • the fitted cap is constructed making use of photogrammetry, computer modeled 3D design methods, 3D printing or a combination thereof, and in some embodiments the fitted cap is constructed to accommodate a specific size and breed of a dog. In some embodiments, the fitted cap is constructed to accommodate a specific canine in an individualized manner.
  • the fitted cap further comprises
  • fNIRS functional nearinfrared spectroscopy
  • infrared sensors may be placed in the sniffing ports, as herein described, while additional sensors, such as IMU, EOG, EMG, fNIRS optodes and others can be incorporated within the fitted caps of this invention and it will be appreciated that any reference to the incorporation of any thus named sensor, will be understood to encompass any commercially known or published description of same or equivalent of same, as the term is an art recognized term.
  • connection refers to, the connection of two components directly, or indirectly, wirelessly or by being physically coupled thereto (e.g., by wire), such that one of the two components receives from the other component either, a signal, a signal comprising data, a command signal, electrical power, or any appropriate connection for completion of the operational task described in connection therewith and including any nuance and embodied aspect of same known to the artisan, as the term is an art recognized term.
  • analog to digital refers to a component which converts an analog signal into a digital signal, and encompasses the myriad of devices and implements to effect same commercially available and published, as will be appreciated by the skilled artisan, as the term is an art recognized term.
  • the present invention relates to a canine brain-computer interface (BCI) signal collection and analysis system, said system comprising:
  • a non-transitory computer-readable storage medium having computer executable instructions stored thereon, which is operationally connected to the receiver and which computer executable instructions when executed by the processor, causes the processor to perform: o acquiring and storing of EEG signals collected by the sensors; and o segmenting the signal sequence into epochs; o generating at least one feature vector for each epoch, wherein each of said feature vectors comprises one or more feature parameters that are associated with a particular characteristic of the respective EEG signal generated from the epoch; o inputting the at least one generated feature vector into a machine learning classifier and applying a preformed classifying model on the at least one generated feature vector to output at least one determination vector for each epoch, wherein the at least one determination vector comprises an indication parameter associating the respective epoch to a canine neurological or behavioral state; and using the at least one determination vector to calculate a defined neurological or behavioral state for the canine.
  • processor refers to a machine-made computer processor commercially available (e.g., by Microsoft) or may refer to the art recognized term.
  • Machine learning classifiers as described herein refer to computer processor based models which enable the development and study of statistical algorithms that can learn from data and generalize to unseen data, or may refer to the art recognized terms.
  • the EEG signals collected are processed and one or more features are extracted from the signals. These features are inputted into a feature vector.
  • the feature vector may be a vector with one or more dimensions that comprise the one or more extracted features. These extracted features may be a function applied to or calculation applied on a respective EEG signal (e.g., a signal in an epoch segment as explained herein).
  • the feature vectors are inputted into the machine learning classifier that applies a preformed classifying model on the feature vectors, and outputs a determination parameter which associates the EEG signals to a certain neurological or behavioral state of the canine.
  • This RMS value of approximately 1.4190 pV would be one element in the feature vector for this 50ms window of EEG data.
  • acquiring and storing of the EEG signals collected by the sensors comprises acquiring signals that were amplified, converted to a digital format, or a combination thereof.
  • the indication parameter indicates the probability of the respective epoch being associated with the canine neurological or behavioral state.
  • said system performance further comprises:
  • said system further comprises a transmitter, which is optionally a wireless transmitter.
  • a mobile device comprises some or all of the components of said system.
  • the processor functions may be performed on a computer cloud system.
  • said processor may subtract signals acquired which represent artifact signals, following independent component analysis (ICA) and Artifact Subspace Reconstruction (ASR).
  • ICA independent component analysis
  • ASR Artifact Subspace Reconstruction
  • the ASR may provide additional clarity and/or idealized results, as same requires tuning to calibration data from the respective individual/species, which in turn may contribute more canine specific information and results.
  • the machine learning classifiers are selected from a group consisting of Support Vector Machine (SVM) and Neural Network (such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs)).
  • SVM Support Vector Machine
  • CNNs Convolutional Neural Networks
  • RNNs Recurrent Neural Networks
  • the invention contemplates making use of xg boost, hidden Markov models, or ensemble approaches, as well, and same are to be considered embodied aspects of the invention.
  • the invention provides a method for determining a neurological or behavioral state of a canine comprising:
  • system acquiring and optionally storing of a series of EEG signals collected by from said canine, whereby said sequence of said series of signals is segmented into one or more signal sequence epochs;
  • each determination vector comprises an indication parameter associating a respective epoch to a pre-determined neurological or behavioral state
  • the final determination is based on a weighted sum of the output determination vectors.
  • Other embodiments may include the final determination being based on a specific value associated with one or more of the output determination vectors (e.g., if a certain value associated with a respective vector exceeds a predetermined threshold, or if at least a predetermined number of vectors contain a value that exceeds a predetermined threshold).
  • Other final determination embodiments may include a combination thereof (e.g., weighted sum combined with one or more vectors having a value that exceeds a predetermined threshold).
  • the indication parameter indicates the probability of the respective epoch being associated with the canine neurological or behavioral state.
  • the method further comprises system establishing defined feature vectors yielding defined determination vectors, which are established to correspond to defined neurological or behavioral states, serving as validation controls of said method.
  • the signals acquired which represent artifact signals are subtracted out from EEG signal sequence epochs or EEG motion signal epochs or a combination thereof.
  • ICA independent component analysis
  • obtaining the series of EEG signals comprises the following steps:
  • the machine learning classifiers are selected from a group consisting of Support Vector Machine (SVM) and Neural Network classifiers (such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs)).
  • SVM Support Vector Machine
  • CNNs Convolutional Neural Networks
  • RNNs Recurrent Neural Networks
  • system segments detected motion signals into epochs, optionally correlating said motion signal epochs with said signal sequence epochs.
  • the method further comprises generating a motion feature vector for each of said motion signal epochs; optionally correlating said motion feature vector with said EEG derived feature vector.
  • each of the motion feature determination vectors comprises an indication parameter associating the motion signal epoch to a pre-determined neurological or behavioral state; and • system calculation of a final determined neurological or behavioral state of the canine based on the output motion feature determination vectors.
  • the method further comprises pre-training a canine to respond to a defined stimulus with a defined neurological or behavioral state, whereby said response results in defined EEG signal stimulation and detection in a controlled setting, and said method results in confirmation of said defined neurological or behavioral state.
  • controlled setting may refer, in some aspects to a laboratory setting, or implementation in the field but with defined and known output in terms of the method so that controlled, accurate pre-training can be achieved.
  • the method comprises having said system acquiring and storing of the EEG signals collected by the sensors comprises acquiring signals that were amplified, converted to a digital format, or a combination thereof.
  • the indication parameter indicates the probability of the respective epoch being associated to the canine neurological or behavioral state.
  • the method further comprises system establishing defined feature vectors yielding defined determination vectors, which are established to correspond to defined neurological or behavioral states, serving as validation controls of said method.
  • the method further comprises pre-training a canine to respond to a defined stimulus with a defined neurological or behavioral state, whereby said response results in defined EEG signal stimulation and detection in a laboratory setting, and said method results in confirmation of said defined neurological or behavioral state.
  • the neurological or behavioral state is the detection of a particular odor and in some embodiments, the neurological or behavioral state is the detection of two or more defined odors and in still other embodiments, the neurological or behavioral state is the detection of a removal of a defined odor from the environment into which the canine is located, which defined odor was previously detected in said environment.
  • the neurological or behavioral state is the detection of a particular sound, or in some embodiments, the neurological or behavioral state is the detection of two or more defined sounds, or, in other embodiments, the neurological or behavioral state is the detection of a removal of a defined sound from the environment into which the canine is located, which defined sound was previously detected in said environment.
  • the neurological or behavioral state is the detection of a particular taste, or in some embodiments, the neurological or behavioral state is the detection of two or more defined tastes, or in still other embodiments, the neurological or behavioral state is the detection of a removal of a defined taste from the environment into which the canine is located, which defined taste was previously detected in said environment.
  • the method further comprises the detection of motion of the canine.
  • motion sensors are operationally connected for acquisition by said system.
  • the method comprises said system segments detected motion signals into epochs, optionally correlating said motion signal epochs with said signal sequence epochs.
  • the method further comprises generating a motion feature vector for each of said motion signal epochs; optionally correlating said motion feature vector with said EEG derived feature vector.
  • the method further comprises:
  • each of the motion feature determination vectors comprises an indication parameter associating the motion signal epoch to a pre-determined neurological or behavioral state
  • the method further comprises higher scoring attribution of said final determined neurological or behavioral state of the canine when said EEG feature determination vectors corelate with said motion feature determination vectors.
  • the devices, system and/or methods of this invention promote determining a neurological or behavioral state of a canine.
  • the devices, system and/or methods of this invention for determining a neurological or behavioral state of a canine may also comprise wherein the neurological or behavioral state correlates with detection of a hazard by the canine.
  • hazard will be understood to any circumstance or object or stimulus that is detectable by the canine, and which would herald an undesirable, or in some aspects, dangerous, or in some embodiments, perilous circumstance or object or stimulus for the canine or as defined by the training program provided to the canine, or for the human aide to, or receiving aid from the canine.
  • the hazard is an explosive material or in some embodiments, the hazard is a physical obstacle.
  • said neurological or behavioral state correlates with detection of a health or disease status in a human subject by the canine and in some embodiments, the neurological or behavioral state correlates with detection relates to the presence of infection, diabetes, cancer, hypoglycemia, narcolepsy, Parkinson’s disease or seizures.
  • the infectious agents may comprise viral or bacterial infection.
  • the fitted cap, system and/or methods of this invention allow for the detection of diseases such as bovine viral diarrhea in cattle, or plant pathogens in agricultural crops.
  • the neurological or behavioral state correlates with detection of fear or stress in a human subject by the canine.
  • said neurological or behavioral state correlates with predictive capability as to whether the canine will serve as a good service dog.
  • said neurological or behavioral state correlates with detection of narcotics by the canine.
  • said neurological or behavioral state correlates with detection of canine health or disease or well-being.
  • signals acquired which represent artifact signals following independent component analysis (ICA) and ASR are subtracted out from EEG signal sequence epochs or EEG motion signal epochs or a combination thereof.
  • ICA independent component analysis
  • obtaining the series of EEG signals comprises the following steps:
  • said transmitting is via wireless transmitter.
  • the machine learning classifiers are selected from a group consisting of Support Vector Machine (SVM) and Neural Network classifiers (such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs)).
  • SVM Support Vector Machine
  • CNNs Convolutional Neural Networks
  • RNNs Recurrent Neural Networks
  • fNIRS sensors are used to collect signals instead of the EEG sensors, mutatis mutandis, and in some embodiments, both are employed cooperatively to glean additional information as herein described.
  • a training stage of generating the preformed classifying model comprising: obtaining training EEG signals comprising EEG signals (as explained herein with optional amplifying, digitalizing, etc.) of a plurality of subjects segmenting the training EEG signals into corresponding training epochs; generating a training feature vector for each training epoch, wherein each of said training feature vectors comprises one or more extracted features (as explained herein, e.g., that are associated with a particular characteristic of the training epoch or that are calculated according to a training signal generated from the training epoch); inputting the generated training feature vectors of each subject into training machine learning classifier along with corresponding true result scores; and generating the preformed classifying model according to said training machine learning classifier.
  • Figure 1 illustrates a System Overview (overall architecture) of a Canine Brain- Computer Interface (BCI) system according to one embodiment of the present invention.
  • BCI Canine Brain- Computer Interface
  • Figure 2 illustrates a 3D Model of a Beagle Generated Using Advanced Portable Imaging Techniques according to an embodiment of the present invention.
  • Figure 3 illustrates the 3D model of the BCI helmet, designed using the 3D head model of the Beagle according to an embodiment of the present invention.
  • Figure 4 illustrates the electrode montages used in non-invasive canine EEG research according to an embodiment of the present invention.
  • Figure 5 A- 5C illustrates the sensor positions within the BCI helmet according to an embodiment of the present invention.
  • Figures 6A-6B illustrate the production of the BCI helmet according to a non-limiting aspect of the invention.
  • Figure 7 shows a non-limiting example of the BCI helmet equipped with an EEG amplifier and the specific electrode montage described in Figure 5.
  • Figure 8 illustrates a non-limiting example of the BCI helmet mounted on the Beagle - Snow - demonstrating the precise and snug fit achieved through custom design and 3D printing.
  • Figure 9 presents visualizations of basehne EEG data collected using the wireless acquisition system integrated into the BCI helmet shown in Figure 8 from Snow, according to a non-limiting embodiment of the present invention.
  • Figure 10 illustrates the correlation between EEG channels, depicting the spatial relationships of the collected baseline data in Figure 9 according to a non-limiting embodiment of the present invention.
  • Figure 11 shows the normalized raw EEG data for all eight channels and their respective power spectral densities (PSDs) calculated at a sampling rate of 1000 Hz according to a non-limiting embodiment of the present invention.
  • PSDs power spectral densities
  • FIG 12 presents a non-limiting example of a screenshot of the graphical user interface (GUI) used for connecting to the EEG system and applying various signal processing algorithms (software functions) to denoise the canine brain data.
  • GUI graphical user interface
  • Figures 13A, 13B and 13C(i), 13C(ii), 13C(iii), 13C(iv) and 13C(v) demonstrate the application of Artifact Subspace Reconstruction (ASR) to the baseline EEG data from Figure 9.
  • ASR Artifact Subspace Reconstruction
  • Figure 14 relates to Body landmark annotation using ML vision models according to a non-limiting embodiment of the present invention.
  • Figures 15A and 15B relate to Computation of quantitative motion data using landmark data according to a non-limiting embodiment of the present invention.
  • Figure 16 illustrates the Integration of quantitative motion data from vision models with EEG data to epoch motion related artifacts according to a non-limiting embodiment of the present invention.
  • Figure 17 relates to an Olfactometer Setup according to a non-limiting embodiment of the present invention.
  • Figures 18A and 18B relate to Event-related potentials for each channel for Scent vs No-Scent Paradigms according to a non-limiting embodiment of the present invention.
  • Figure 19 relates to a Comparison of summed-up Event-Related Potential across channels for each paradigm according to a non-limiting embodiment of the present invention.
  • Figure 20 relates to Support Vector Machine trained on ERP data across paradigms according to a non-limiting embodiment of the present invention.
  • Figure 21 relates to Improvement in classification score with increase in training data according to a non-limiting embodiment of the present invention.
  • Figure 22A shows a Canine Sniff Station Setup according to a non-limiting embodiment of the present invention.
  • Figure 22B shows an ERP Analysis Sniff Station Data according to a non-limiting embodiment of the present invention.
  • Figure 23(i) illustrates the integration of infrared (IR) sensor data with EEG recordings to precisely demarcate canine sniffing events, enhancing olfactory response analysis in the BCI system according to a non-limiting embodiment of the present invention.
  • IR infrared
  • Figure 23(ii) shows the averaged EEG channel signals for both target (gunpowder) and non-target (tissue paper) sniffs according to a non-limiting embodiment of the present invention.
  • Figure 23(iii) shows Time-Frequency Representation for Target and Non-target according to a non-limiting embodiment of the present invention.
  • Figure 23(iv) shows Time-Frequency Representation (Target - Non-target) according to a non-limiting embodiment of the present invention.
  • Figure 23(v) shows PSD of Averaged Signals for Target and Non-target Sniffs (1-40 Hz) according to a non-limiting embodiment of the present invention.
  • Figures 24A, 24B, 24C, 24D(i), 24D(ii) and 24D(iii) relate to Classification of Sniff Station EEG data using Support vector machines according to a non-limiting embodiment of the present invention.
  • This invention provides, in some embodiments, this invention provides a customized brain-computer interface (BCI) cap for canines, systems comprising same and methods of use of the BCI caps as herein described. It will be appreciated that in any embodiment as described herein for the BCI caps, that the systems and methods may make use of same.
  • BCI brain-computer interface
  • Photogrammetry A detailed 3D model of the dog’s head is generated using methods such as methods that include a smartphone camera with a dedicated app or a specialized photogrammetry setup. This model captures the precise contours and dimensions of the canine skull, accommodating variations across different breeds and individual dogs.
  • Wireless Amplifier System according to a non-limiting aspect of the invention:
  • the present invention system incorporates advanced acquisition technologies to wirelessly collect high-quality brain data from canines. This component ensures that the data collection process is efficient, unobtrusive, and suitable for various environments and scenarios.
  • Wireless Connectivity The system utilizes advanced wireless technology to transmit EEG data from the helmet to a receiving device, such as a computer or mobile device. This setup allows for free movement, enabling data collection in naturalistic settings without the constraints of wired connections. Multiple wireless technologies can be employed to achieve this, ensuring flexibility and reliabihty.
  • Amplifier Systems The system is compatible with various amplifier types, including off-the- shelf amplifiers recording at various sampling rates including at 1kHz. This flexibility allows for different configurations based on specific needs and preferences.
  • the amplifiers ensure that the signals are amplified and digitized accurately for subsequent processing, accommodating a range of acquisition capabilities.
  • Data Synchronization The system includes robust software for timeline synchronization and data integration. This ensures that the collected brain data is accurately aligned with experimental events and other physiological measurements. Various synchronization methods and software solutions can be employed to achieve precise data integration.
  • the system employs advanced signal processing techniques to denoise and enhance the quality of the collected canine brain data. These techniques are specifically tuned to address the unique challenges associated with canine EEG.
  • Bandpass Filtering and Normalization According to one embodiment, the data undergoes bandpass filtering to remove frequencies outside the range of interest, reducing noise and artifacts. Various filtering techniques can be employed to achieve this, depending on the specific requirements of the study. Normalization techniques are applied to standardize the data, making it suitable for further analysis. Different normalization methods can be used based on the nature of the data and the desired outcome.
  • ICA Independent Component Analysis
  • ASR Artifact Subspace Reconstruction
  • Automatic Artifact Identification Algorithms The system employs automatic artifact identification algorithms that utilize data from Electromyography (EMG), Electrooculography (EOG), and motion sensors. These algorithms detect and correct artifacts, ensuring high-quality EEG data. Various automatic identification methods can be used to address different types of artifacts, enhancing the overall reliability of the data.
  • EMG Electromyography
  • EOG Electrooculography
  • motion sensors These algorithms detect and correct artifacts, ensuring high-quality EEG data.
  • Various automatic identification methods can be used to address different types of artifacts, enhancing the overall reliability of the data.
  • GUI Graphical User Interface
  • the system leverages advanced machine learning models to analyze the processed EEG data and provide predictive insights into canine cognition and behavior.
  • Feature Extraction Key features are extracted from the processed EEG data using both timedomain and frequency-domain analyses. These features serve as inputs for the machine learning models. Various feature extraction methods can be utilized depending on the specific requirements of the study.
  • Machine learning algorithms such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs), are trained to predict various cognitive and behavioral states of canines. These models can identify patterns associated with specific tasks, behavioral states, emotions, or cognitive processes. Different machine learning models can be employed based on the nature of the data and the desired predictions.
  • CNNs Convolutional Neural Networks
  • RNNs Recurrent Neural Networks
  • Real-Time Analysis The system supports real-time data analysis, enabling immediate feedback and intervention based on the predicted states. This capability is particularly useful in operational settings, such as detection tasks or training sessions.
  • the system's real-time analysis can be customized to suit various operational needs, providing flexibility in its applications.
  • Figure 2 illustrates a 3D Model of a Beagle Generated Using Advanced Portable Imaging Techniques.
  • the 3D model of the Beagle’s head in Figure 2 is generated using advanced imaging techniques that ensure a precise fit for the BCI helmet.
  • the figure in question was produced to capture the head model of a 3 year-old female Beagle called Snow lying down on a mattress in an office in South Bangalore in March 2024.
  • One non-limiting method to produce this model involves using photogrammetry, where multiple photographs of the Beagle’s head are captured from different angles and processed to create a detailed 3D model. This process involves feature detection and matching algorithms that align and stitch the images together, creating a point cloud that accurately represents the surface geometry of the Beagle's head.
  • structured light scanning and depth-sensing technologies such as those found in modern smartphones with TrueDepth cameras, can be used. These systems project a pattern of infrared light onto the subject and capture the distortion of this pattern to generate a depth map. The depth data is combined with RGB images to construct a comprehensive 3D model. This video-based approach allows for real-time depth capture as the device is moved around the subject, integrating depth information and color data to produce a highly accurate 3D mesh.
  • Figure 3 illustrates the 3D model of the BCI helmet, designed using the 3D head model of the Beagle. Creating this head mesh involves capturing precise contours and dimensions of the canine skull through advanced imaging techniques. The head mesh undergoes detailed editing to refine and optimize the design. This particular non-limiting example model was refined in Nomad Sculpt - a polygon modeling software.
  • the technical process includes:
  • the refined head mesh is used as a base for designing the BCI helmet.
  • the design incorporates ergonomic principles to ensure it conforms to the natural contours of the head. Key elements include shaping the helmet to fit the head, adding padding for comfort, and designing adjustable components for a secure fit.
  • ACI Animal- Computer Interaction
  • figure 4 illustrates the electrode montages used in non-invasive canine EEG research, highlighting how these arrangements are adapted for the BCI helmet design.
  • the present invention embodiment design tunes the electrode placements to ensure optimal signal quality.
  • Traditional human EEG montages like the 10-20 system, are adapted to account for the significant variance in head shape and size across dog breeds.
  • the helmet design incorporates grooves and channels that securely hold electrodes in positions validated by prior research, minimizing muscular artifacts and ensuring reliable contact with the scalp.
  • the reference electrode is strategically positioned to enhance signal accuracy, considering anatomical landmarks that provide stable and artifact-free readings. This tailored approach ensures that the helmet accommodates the unique anatomical features of canines while maintaining high-quality EEG data collection.
  • figure 5 A illustrates the sensor positions within the BCI helmet, focusing on the design elements that secure the EEG electrodes.
  • the helmet features precisely designed grooves and channels that accommodate various electrode types, ensuring they remain securely in place. These grooves prevent electrode movement, enhancing signal quality by maintaining consistent and close contact with the scalp.
  • the structured layout of the grooves aligns with anatomical landmarks to capture accurate EEG data. Additionally, the design applies gentle pressure to keep the electrodes firmly against the skin, minimizing artifacts and ensuring reliable signal acquisition.
  • the grooves and channels are adaptable, allowing for different electrode configurations to suit various research needs.
  • midline electrodes of Fz Cz and Pz are strategically placed on the sagittal crest of the dog’s head, which provides a bony structure that minimizes muscle artifacts.
  • Fz is designed to correspond to the midline of the dog’s eyebrows which is called the glabella.
  • Pz is designed to be on the bony knob of the dog’s sagittal crest, which is called the external occipital protuberance.
  • Cz is designed to be exactly in between Fz and Cz.
  • Fpl and Fp2 are designed to be equidistant from and in the same plane as Fz.
  • T3 and T4 are designed to form an equilateral triangle between Fz and Cz.
  • P3 and P4 are designed to form an equilateral triangle between Cz and Pz.
  • This is the montage represented in Figure 5A.
  • Figure 5B shows the position of the montage on Snow herself.
  • Figure 5C showcases the grooves on the helmet in the final printable version according to one embodiment of the present invention.
  • An example montage represents a system with 9 electrodes and 8 channels.
  • An electrode represents an electrical voltage measurement sensor, a channel represents an active source of signal.
  • the below system uses a single reference electrode, to compute the signal values in the remaining 8-ch example system.
  • the reference electrode is positioned at Pz, to minimize the disruption in signal as it is placed on the external occipital protuberance.
  • the montage may be tuned to specific purposes and does not have to be the same as the below.
  • the montage may consist of more electrodes than the 9 in the below representation. It may utilize a different electrode as the reference. It may also use electrodes in the occipital region (03, 04) instead of (P3, P4) to capture information from visual processing.
  • the same methods of grounding in anatomical markers, and replication in a reproducible way may be used to generate a bespoke montage fit for the specific use-cases.
  • FIGs 6A-6B illustrate the production of the BCI helmet according to a non-limiting aspect of the invention, using 3D printing technology.
  • the helmet is printed using materials such as Thermoplastic Polyurethane (TPU) and Poly lactic Acid (PLA), which provide the necessary durability and flexibility.
  • TPU Thermoplastic Polyurethane
  • PLA Poly lactic Acid
  • the 3D printing process allows for precise customization, ensuring a snug fit that matches the specific anatomical features of the dog’s head.
  • the design includes grooves and connectors for wiring, securing the electrodes and maintaining reliable contact with the scalp.
  • 3D printing is the primary method used, other embodiments may include other manufacturing techniques such as injection molding or thermoforming can also be employed to achieve similar precision and customization. This approach ensures the helmet is both functional and comfortable, tailored to the unique needs of different canine breeds.
  • Nozzle Size 0.4 mm
  • volumetric speed limitation o Max volumetric speed: 12 mm A 3/s
  • Model Preparation The helmet was designed using polygon modeling and CAD software, incorporating all necessary slots and compartments for electrodes and electronics. The model was optimized for 3D printing with considerations for support and overhangs.
  • Slicing The model was loaded into slicing software compatible with the Bambu Lab PIS, applying the specified print settings. Slicing parameters such as layer height, infill density, and support settings were correctly configured.
  • Post-Processing o
  • the printed helmet was carefully removed from the build plate. o Any support structures were removed. o Rough edges or surfaces were smoothed if necessary, using fine sandpaper or a similar tool. o The helmet was test-fitted on the dog and any necessary adjustments were made.
  • Figure 7 shows a non-limiting example of the BCI helmet equipped with an EEG amplifier and the specific electrode montage described in Figure 5.
  • the design ensures that electrodes maintain consistent and close contact with the scalp, enhancing signal quality. Grooves and connectors secure the wiring, preventing movement and minimizing artifacts. This configuration allows for easy adjustment of electrodes and supports various EEG sensor types, providing reliable and high-quality data collection during different activities.
  • the canine Brain-Computer Interface (BCI) system utilizes sintered AglAgCl wet electrodes in conjunction with an off-the-shelf 8-channel EEG amplifier.
  • the system's design accommodates a diverse array of electrode types and amplifier configurations, ensuring adaptability and optimization for various research or clinical scenarios.
  • Alternative electrode options include dry electrodes constructed from materials such as conductive polymers, carbon-based composites, or metal alloys, each offering distinct advantages in terms of signal quality, comfort, and long-term stability.
  • the system supports the integration of active wet or dry electrodes, which incorporate pre-amplification circuitry within the electrode housing, thereby enhancing signal-to-noise ratio and reducing motion artifacts.
  • the amplifier subsystem while fundamentally similar to human BCI amplifiers, is specifically engineered to meet the unique requirements of canine subjects across various breeds and sizes.
  • the primary design constraint is the minimization of size and weight to ensure comfortable attachment and unimpeded mobility for canines of all dimensions.
  • the amplifier is conceptualized as a centralized unit comprising five sub-components: (1) An amplifier circuit for signal enhancement, capable of boosting weak neural signals to levels suitable for digital conversion; (2) An analog-to-digital converter (ADC), which can be implemented using high-performance chips such as the ADS 1299 Low-Noise 24-Bit ADC specifically designed for biopotential measurements, offering 4, 6, or 8 channel configurations; (3) A signal processing system built around either a microcontroller or microprocessor, such as the ESP32-C3, executing algorithms for filtering and feature extraction relevant to the targeted cognitive states; (4) A wireless transmitter for real-time data broadcasting to external receivers, facilitating immediate monitoring and analysis; and (5) A power source, typically a lightweight, high-capacity battery, ensuring extended operational autonomy.
  • a power source typically a lightweight, high-capacity battery, ensuring extended operational autonomy.
  • the system's architecture allows for the integration of alternative components with similar specifications, providing flexibility in research and development contexts.
  • This modular design enables seamless integration with both commercial off-the-shelf amplifiers and custom-designed solutions, maintaining consistent performance across various hardware configurations while adapting to evolving technological advancements in the field of bio-signal acquisition and processing.
  • Figure 8 illustrates a non-limiting example of the BCI helmet mounted on the Beagle - Snow - demonstrating the precise and snug fit achieved through custom design and 3D printing.
  • the helmet's ergonomic design ensures it remains securely in place, allowing for consistent and reHable wireless EEG recording.
  • Adjustable straps are included to further secure the helmet, ensuring it fits comfortably and stays in place during the dog’s movement. Multiple types of straps can be used, that include padding and different designs to minimize discomfort. This setup enables the dog to move freely while maintaining high-quality data acquisition, sselling the helmet's effectiveness for wearable and wireless EEG monitoring in naturalistic settings.
  • Figure 9 presents visualizations of baseline EEG data collected using the wireless acquisition system integrated into the BCI helmet shown in Figure 8 from Snow, according to a non-limiting embodiment of the present invention.
  • Baseline data was collected during rest, during no stimulus or activity. Data collection occurred in a noise-free and sealed room, with Snow lying down on a mattress with only two experimenters present. Data visualized is after initial helmet mounting and 10 minutes of settling down using positive reinforcement.
  • FIG. 10 illustrates the correlation between EEG channels, depicting the spatial relationships of the collected baseline data in Figure 9 according to a non-limiting embodiment of the present invention. Pearson correlation coefficient was calculated between all the pairs of channels. Pearson correlation is a statistical measure that quantifies the strength and direction of a linear relationship between two continuous variables.
  • R It is denoted by R and ranges from - 1 to 1.
  • the figure shows a correlation matrix where each cell represents the Pearson correlation coefficient between pairs of channels. High positive correlation values (close to 1 ) indicate good phase and amplitude synchronous neural activity i Values close to zero indicate no covariant relationship between two channels. And high negative correlations indicate synchronous but asymmetric neural activity.
  • This analysis verifies that neural signals captured by the electrodes are in check with the montage used, effectively capturing localized brain activity. The meaningful nature of this data is evidenced by higher correlations observed in the frontal regions, indicating synchronized neural activities in those areas.
  • the spatial correlation matrix confirms the accurate reflection of neural dynamics across different brain areas, ensuring the integrity of the data collection process and identification of artifacts.
  • FIG 11 shows the normalized raw EEG data for all eight channels and their respective power spectral densities (PSDs) calculated at a sampling rate of 1000 Hz according to a non-limiting embodiment of the present invention.
  • PSDs power spectral densities
  • the PSDs calculated using Welch's method, provide insights into the frequency components of the EEG signals by illustrating the power distribution across different frequencies.
  • Welch’s method reduces the variance of spectral estimates by segmenting the signal, windowing each segment, applying the Fast Fourier Transform (FFT), and averaging the resulting periodograms.
  • the FFT efficiently converts timedomain signals into their frequency-domain representations, allowing for detailed frequency analysis.
  • PSD plots reveal the expected decrease in power with increasing frequency and show harmonic peaks at 50 Hz, indicating powerline noise that needs removal. Analyzing the PSDs helps assess data quality by identifying characteristic neural oscillations (delta, theta, alpha, beta, gamma) and understanding the spatial orientation of frequency and power distribution across channels. In preferred embodiments, this detailed spectral analysis is crucial for understanding the underlying neural dynamics and ensuring high-quality data collection.
  • FIG 12 presents a non-limiting example of a screenshot of the graphical user interface (GUI) used for connecting to the EEG system and applying various signal processing algorithms (software functions) to denoise the canine brain data.
  • GUI graphical user interface
  • the GUI facilitates real-time data monitoring and control, allowing researchers to select and apply a range of signal processing algorithms, including bandpass filtering, Independent Component Analysis (ICA), Artifact Subspace Reconstruction (ASR), and automatic artifact identification using EMG, EOG, and motion data.
  • ICA Independent Component Analysis
  • ASR Artifact Subspace Reconstruction
  • Users can customize the processing pipeline with specific parameters and settings, while visualization tools display raw and processed EEG data, correlation matrices, and power spectral densities. This user-friendly interface ensures efficient preprocessing and high- quality data for advanced research.
  • Figure 12 presents the baseline data from Figure 9, using various filters, and other preprocessing non-limiting methods.
  • Load Data Allows the user to import EEG data for analysis.
  • Notch Frequency This removes electrical noise at this specific frequency through a targeted frequency rejection process, particularly the 60 Hz (or 50 Hz in some countries) interference from power lines.
  • a notch filter is a type of band-stop filter, which is a filter that attenuates frequencies within a specific range while passing all other frequencies unaltered. For a notch filter, this range of frequencies is very narrow. It creates a sharp "notch" in the frequency response, dramatically reducing the amplitude of signals at the target frequency while minimally affecting nearby frequencies.
  • Quality Factor This determines the width and sharpness of the notch filter. A higher quality factor creates a narrower, more targeted notch.
  • PCA Components This determines the number of principal components for PCA analysis. PCA is used to reduce the dimensionality of data while retaining most of the variation in the dataset. More components retain more of the original signal's complexity but may include noise while fewer components simplify the data more, potentially making patterns clearer but at the risk of losing subtle information.
  • ICA Components This sets the number of independent components for ICA analysis. ICA separates a multivariate signal into additive, statistically independent components. It assumes that the EEG signal is a mix of independent sources and tries to unmix these sources. Each component ideally represents an independent source of brain activity or artifact (like eye blinks or muscle movement). More components can separate more distinct sources of activity but risk over-splitting the signal while fewer components provide a more generalized separation but might group distinct sources together.
  • Y Scale (uV) This defines the vertical scale for plotting in microvolts.
  • Time Resolution This sets the time window for analysis.
  • Raw Normalized Time Series This would display the original EEG signal data after normalization, showing how the electrical activity varies over time without any filtering applied.
  • Notch + FIR Filtered PSD This shows the Power Spectral Density (PSD) of the signal after applying both Notch and Finite Impulse Response (FIR) filters.
  • the FIR filter is employed as a bandpass filter, specifically designed to isolate and preserve frequencies within a defined range while attenuating frequencies outside this band. It allows one to focus on the frequency bands of interest in EEG analysis (e.g., delta, theta, alpha, beta), which typically fall between 0.5 Hz and 40 Hz. By excluding frequencies outside the specified band, the filter effectively removes various sources of noise, including high-frequency artifacts and low-frequency drift.
  • the PSD represents the strength of signal variations as a function of frequency.
  • Raw PSD The Power Spectral Density of the raw, unfiltered signal, allowing comparison with the filtered version.
  • Notch + FIR Filtered Time Series This displays the time series data after applying both notch and FIR filters, which remove specific frequency noise and smooth the signal.
  • PCA Principal Component Analysis
  • ICA Independent Component Analysis
  • PSD The Power Spectral Density of the processed signal (after all selected filters and analyses).
  • ICA Topo Map This would display a topographical map of the scalp showing the spatial distribution of the independent components identified by ICA.
  • the "Same Axis" option for multiple visualizations can overlay different representations of the data on a single graph, allowing for direct comparison.
  • Figures 13A-13E demonstrate the application of Artifact Subspace Reconstruction (ASR) to the baseline EEG data from Figure 9, highlighting its effectiveness compared to standard filtering methods.
  • ASR Artifact Subspace Reconstruction
  • rASR Riemannian Artifact Subspace Reconstruction
  • ASR Algorithm according to a non-limiting aspect of the present invention:
  • ASR Artifact Subspace Reconstruction
  • Calibration Phase o Purpose of Calibration Data: Calibration data is essential as it provides a reference for the statistical properties of clean EEG signals. This data, typically at least 1 minute long, should be recorded from the dog at rest under conditions similar to the actual recording. o Filtering and Segmentation: The calibration EEG data is filtered and segmented into overlapping windows of length w with overlap o. For each window Xi, the sample covariance matrix Ci is computed as:
  • T p + k * o where k is a tuning parameter.
  • o SPD matrices such as covariance matrices in EEG data, lie on a Riemannian manifold — a curved space where traditional Euclidean geometry is not applicable. In this context, distances are computed along curved geodesics rather than straight lines.
  • This geometry- aware approach has several advantages: o 1. Improved Averaging: The Riemannian center of mass provides a more accurate average, reducing the swelling effect observed with Euclidean means. o 2. Enhanced Variance Preservation: PGA maximizes variance preservation in the curved space, leading to more precise artifact detection and removal. o 3. Robust Covariance Estimation: By focusing on the geometry of SPD matrices, rASR provides a robust and unbiased covariance estimate, improving the overall quality of EEG data reconstruction.
  • FIG. 13A ASR Corrected EEG Time series vs. Filtered EEG Time series (0.5-40 Hz)
  • This subplot compares a time window of the ASR corrected EEG signal (solid blue) to the pre ASR corrected EEG data (dotted red) for channel T3. This data is collected from a freely moving dog, sniffing odour samples. It is quite evident from the plot that ASR very effectively corrects for artifacts introduced into the EEG signal, while also preserving the shape and dynamics of the original signal.
  • This subplot shows pairwise correlations between the 8 EEG channels for ASR corrected and pre ASR filtered (0.5-40 Hz) data.
  • the plot on the right is for ASR and the left is pre ASR.
  • the unrealisticly high correlation between channels in the pre-ASR corrected data is indicative of the large amount of noise represented in the time series of each channel, which leads to the high correlation values.
  • the plot on the right has a much more reasonable correlation structure, and the true correlation of neural activity is enhanced. This highhghts the performance and significance of the ASR algorithm.
  • the figures present a comparison of EEG time series data across Alpha, Beta, Delta, Gamma, and Theta bands, before and after Artifact Subspace Reconstruction (ASR) correction.
  • Pre-ASR data is shown with red dotted lines, while ASR corrected data is in blue solid lines.
  • Each figure demonstrates the ASR algorithm's effectiveness in reducing movement-related artifacts.
  • artifacts from eye movements and posture changes are reduced.
  • muscle artifacts from sniffing show significant reduction.
  • the Delta band illustrates diminished slow wave artifacts from head and body movements.
  • the Gamma band highlights decreased high-frequency muscle artifacts from rapid movements.
  • the Theta band shows reduced artifacts from rhythmic head movements and walking.
  • the ASR correction effectively cleans the EEG signals, making them more interpretable by mitigating movement and muscle-related artifacts associated with a dog moving and sniffing.
  • ASR's application provides effective artifact correction, maintaining high-quality EEG data essential for accurate neural activity analysis in canine studies.
  • Figure 14 relates to Body landmark annotation using ML vision models according to a non-limiting embodiment of the present invention. More particularly, figure 14 illustrates the use of DeepLabCut (DLC), an advanced machine learning (ML) vision model, to automatically annotate body landmarks on a Beagle using camera data.
  • DLC employs a deep neural network to detect and track 23 key body points such as eyes, nose, joints, and more with high precision.
  • the model is trained on annotated images to learn the spatial relationships and visual features associated with each landmark, enabling it to generalize and accurately identify these points across new images and video frames. This capability allows for precise monitoring of the dog's posture and movements, facilitating the study of behavior and physiological states. Additionally, the system can estimate more landmarks if needed, providing flexibility and detailed analysis options.
  • CNNs convolutional neural networks
  • Figures 15A and 15B relate to Computation of quantitative motion data using landmark data according to a non-limiting embodiment of the present invention. More particularly, Figures 15A and 15B demonstrate the process of computing quantitative motion data from body landmark annotations obtained using machine learning vision models like DeepLabCut.
  • the top graph (15A) shows the loaded position data for various landmarks, plotted against frame numbers, including x and y coordinates along with the likelihood of the detected points.
  • the bottom graph (15B) represents the computed velocity of the motion derived from the position data over time.
  • the system calculates velocity, providing insights into the speed and direction of movement. This process involves differentiating the positional data to obtain velocity values, allowing for detailed analysis of the dog's movement dynamics. This method can be applied to each landmark, facilitating comprehensive motion analysis in behavioral studies.
  • Figure 16 illustrates the Integration of quantitative motion data from vision models with EEG data to epoch motion related artifacts according to a non-limiting embodiment of the present invention.
  • Figure 16 demonstrates the integration of quantitative motion data obtained from machine learning vision models with EEG data to identify and epoch motion-related artifacts.
  • the EEG channels (Fz, Cz, T3, T4, P3, P4, Fpl, Fp2) show raw signals plotted over time.
  • the bottom graphs represent the x and y positional data derived from the tracked landmarks.
  • This integration enables the creation of epochs that correspond to specific motion events, allowing for the isolation and analysis of neural data during periods of reduced motion interference. This process is crucial for improving the accuracy of EEG data interpretation by accounting for and mitigating the impact of motion-related artifacts.
  • the epochs identified in Figure 16 are used to train an automatic motion artifact identification model from EEG data.
  • the process begins with the synchronized EEG and motion data, where epochs are labeled to indicate periods affected by motion-related artifacts.
  • This labeled data serves as training input for a machine learning model, typically a neural network.
  • the model is trained to recognize patterns in the EEG signals that correspond to motion artifacts by learning from the labeled epochs. Once trained, the model can automatically identify and label motion artifacts in new EEG data, enhancing the accuracy of EEG analysis by isolating and mitigating the impact of these artifacts. This approach leverages the precise motion data to improve the reliability and utility of EEG recordings in behavioral and neurological studies.
  • EEG data can provide predictive inferences on various cognitive and behavioral states of canines.
  • the objective is to distinguish between periods when a dog is exposed to a specific scent versus no scent.
  • This paradigm involves presenting controlled olfactory stimuli to the dog and recording the corresponding EEG signals using a wireless BCI helmet.
  • the EEG data is then segmented into epochs based on the presence or absence of the scent, forming the basis for further analysis.
  • Figure 17 relates to an Olfactometer Setup according to a non-limiting embodiment of the present invention.
  • An in-house olfactometer was designed and built. The device allows for the precise delivery of odors according to a computer script. The canine participant is trained to keep focused on the funnel, through which the scent is delivered by the opening of a valve connected to tubing to an air compressor.
  • FIGs 18A and 18B relate to Event-related potentials for each channel for Scent vs No-Scent Paradigms according to a non-limiting embodiment of the present invention.
  • Event- Related Potentials are a type of brainwave response that are time-locked to specific sensory, cognitive, or motor events. They are extracted from the EEG data by averaging the brain's electrical activity across multiple occurrences of the same event, thereby isolating the neural response to that event from the background EEG noise. ERPs are particularly useful for understanding the timing and functional aspects of cognitive processes because they provide detailed temporal information about the brain's response to specific stimuli.
  • Figures 18A and 18B illustrate the ERPs recorded from eight EEG channels during the presentation of a scent ( Figure 18 A) and no scent ( Figure 18B) in the olfactometer paradigm.
  • Each graph represents the averaged electrical activity from multiple trials, highlighting the distinct neural patterns elicited by the presence and absence of the olfactory stimulus.
  • the first 2.5 seconds after a valve connected to a chamber with gunpowder (Scent) vs the first 2.5 seconds after a valve connected to an empty chamber (No Scent) is plotted.
  • 40 trials with Scent and 40 trials of No Scent were used and averaged to create the ERP waves above for each channel from a single Beagle Snow.
  • the ERPs for the scent condition show a clear response, with higher amplitude peaks and distinct waveforms, compared to the no-scent condition. This difference in neural response indicates the brain's ability to detect and process olfactory stimuli, providing insights into canine cognition and sensory processing.
  • These ERP patterns can serve as features for machine learning models to predict the presence of specific scents based on EEG data, demonstrating the potential of combining neurotechnology with olfactometry for advanced cognitive and behavioral research in canines.
  • Figure 19 relates to a Comparison of summed-up Event-Related Potential across channels for each paradigm according to a non-limiting embodiment of the present invention. More particularly, Figure 19 presents a comparison of the summed-up Event-Related Potentials (ERPs) across multiple channels for each olfactometer paradigm.
  • the graph illustrates the averaged electrical activity from EEG channels Fz, Cz, T3, T4, P3, P4, Fpl, and Fp2, providing a comprehensive view of the neural response to the presence (blue) and absence (orange) of a scent.
  • the summed ERPs highlight the differences in brain activity between the two conditions.
  • the presence of a scent elicits more pronounced and distinct neural responses, with noticeable peaks and troughs at specific time intervals, compared to the no-scent condition.
  • These differences in the ERP waveforms reflect the brain's processing of olfactory stimuli and underscore the potential of using ERPs as biomarkers for cognitive and sensory states in canines.
  • machine learning models can be trained to predict the presence of specific scents, offering valuable insights into canine cognition and olfactory processing. This approach demonstrates the efficacy of combining EEG data with advanced analytical techniques to decode neural responses and enhance our understanding of canine behavior and sensory perception.
  • Figure 20 relates to Support Vector Machine trained on ERP data across paradigms according to a non-limiting embodiment of the present invention. It is noted that Same features from Figure 24 (hereinafter) are used here as well.
  • Figure 20 displays the confusion matrix for a Support Vector Machine (SVM) model trained on ERP data across the olfactometer paradigms. This matrix provides a visual representation of the classification performance of the SVM in predicting the presence or absence of a scent based on EEG data.
  • SVM Support Vector Machine
  • the confusion matrix includes four quadrants that summarize the true positive, true negative, false positive, and false negative rates of the model.
  • the top-left quadrant (44) indicates the number of correct predictions for the no-scent condition, while the bottom-right quadrant (45) shows the correct predictions for the scent condition.
  • the top-right (19) and bottom-left (21) quadrants represent the misclassifications, where the model incorrectly predicted the presence or absence of a scent.
  • the model's performance metrics such as precision, recall, and Fl -score, can be derived from this matrix. These metrics are crucial for evaluating the effectiveness of the SVM in distinguishing between the two conditions based on ERP features.
  • the results highlight the potential of machine learning models to leverage ERP data for predicting cognitive and sensory states, advancing the understanding of canine olfactory processing and enhancing the accuracy of behavioral predictions.
  • Figure 21 relates to Improvement in classification score with increase in training data according to a non-limiting embodiment of the present invention. More particularly, Figure 21 illustrates the enhancement in classification performance as the amount of training data increases for the Support Vector Machine (SVM) model.
  • SVM Support Vector Machine
  • the learning curve graph presents both the training score and the cross-validation score relative to the number of training examples.
  • the red line represents the training score, indicating how well the model fits the training data.
  • the green line shows the cross-validation score, which estimates the model's performance on unseen data. As the number of training examples grows, the cross-validation score consistently improves, reflecting better generalization and predictive accuracy of the model.
  • FIG 22A shows a Canine Sniff Station Setup according to a non-limiting embodiment of the present invention.
  • the Sniff Station paradigm involves a dog freely exploring an area while sniffing both target and control samples, with EEG data being recorded simultaneously.
  • EEG is captured from 8 channels (Fz, Cz, T3, T4, P3, P4, Fpl, Fp2) at a 1000 Hz sampling rate.
  • This novel data collection method integrates multiple sensors, including EEG, high frequency infrared (IR) sensors, and accelerometers, all acquired synchronized using the Lab Streaming Layer (LSL).
  • Video data is recorded at 30 Hz and post-processed with DeepLabCut (DLC) to extract coordinates of key landmarks.
  • DLC DeepLabCut
  • the timing of sniffs is precisely calculated using a combination of IR sensor data, a post-processing filtering algorithm for IR signals, accelerometer data, and the DLC-derived coordinates of the dog’s nose. This multimodal approach ensures robust and precise data acquisition, facilitating downstream classification and exploratory analyses.
  • FIG 22B shows an ERP Analysis Sniff Station Data according to a non-limiting embodiment of the present invention.
  • the data processed for this analysis comprised five sessions from the dog named Snow.
  • the dataset included 447 no sniff epochs (moving between samples, eating reward) with an average duration of 32.96 seconds, 222 non-target sniff epochs (tissue paper) with an average duration of 0.40 seconds, and 220 target sniff epochs (gunpowder) with an average duration of 0.43 seconds.
  • the Event-Related Potential (ERP) for target and non-target sniffs was calculated to analyze the neural responses.
  • ERP Event-Related Potential
  • ERP Event-Related Potential
  • ERPs are important because they provide insights into the timing and neural mechanisms underlying sensory and cognitive processes. By analyzing ERPs, researchers can understand how the brain processes different stimuli, how quickly these processes occur, and what neural pathways are involved. In this study, ERPs help in identifying how the dog's brain distinguishes between target and non-target odors, providing valuable information on olfactory processing and the potential for training dogs in odor detection tasks.
  • the non-limiting example ERPs calculated here are averaged across all the epochs that have a length of at least 250 ms. For epochs longer than 250 ms, the data is truncated to the first 250 ms. This approach is used to understand the nature of the onset of sniff representation neurally for both target and control odors.
  • Figure 23(i) illustrates the integration of infrared (IR) sensor data with EEG recordings to precisely demarcate canine sniffing events, enhancing olfactory response analysis in the BCI system according to a non-limiting embodiment of the present invention.
  • the image displays a continuous EEG waveform across multiple channels, temporally aligned with IR sensor data represented as discrete events or peaks.
  • Superimposed on the EEG trace are green-highlighted windows indicating general sniffing events, triggered by IR sensor activations that detect the proximity of the dog's nose to odor samples. Within these green windows, red-highlighted regions denote interactions with positive (target) odor samples, facilitating differentiation between responses to target and non-target odors.
  • Figure 23(iii) shows Time-Frequency Representation for Target and Non-target according to a non-limiting embodiment of the present invention. These plots compare the timefrequency representations for target (left) and non-target (right) sniffs. The color indicates the power at different frequencies over time.
  • the target sniff (gunpowder) shows higher power in certain frequency bands compared to the non-target sniff (tissue paper), indicating specific neural activity patterns associated with detecting the target odor.
  • Figure 23(iv) shows Time-Frequency Representation (Target - Non-target) according to a non-limiting embodiment of the present invention.
  • This plot illustrates the difference in power across frequencies and time between target and non-target sniffs. It highlights specific frequency bands where differences are more pronounced.
  • the time-frequency representation is calculated using wavelet transforms or short-time Fourier transforms to decompose the signal into its frequency components over time. This analysis is crucial for understanding how the neural response to odors evolves over time and identifies frequency bands that are particularly responsive to target odors. The significant differences in certain bands suggest that these frequencies are critical for the dog's recognition and processing of the target odor.
  • FIG. 23(v) shows PSD of Averaged Signals for Target and Non-target Sniffs (1-40 Hz) according to a non-limiting embodiment of the present invention.
  • the power spectral density (PSD) plot shows the distribution of power across different frequencies for target and non-target sniffs. Solid lines represent target sniffs, while dotted lines represent non-target sniffs. Higher power is observed in lower frequencies for target sniffs, indicating more pronounced neural oscillations in those frequency bands when detecting the target odor. This implies that certain frequency bands are more active and possibly more critical for the neural processes involved in odor detection.
  • Figure 23(i) shows a grand average over all the channels for target (in solid) and non target (dotted).
  • Figures 24A-24D relate to Classification of Sniff Station EEG data using Support vector machines according to a non-limiting embodiment of the present invention. These figures present the classification results of EEG data collected during the above described Sniff Station paradigm. The classification was performed using a Support Vector Machine (SVM) classifier.
  • SVM Support Vector Machine
  • Support Vector Machines are a powerful supervised learning model used for classification and regression tasks.
  • the fundamental idea behind SVM is to find a hyperplane that best divides a dataset into classes.
  • SVM aims to find the optimal hyperplane that maximizes the margin between different classes.
  • the margin is defined as the distance between the hyperplane and the nearest data point of each class.
  • This optimal hyperplane ensures that the classification model has the best generalization ability.
  • the optimization problem for finding the hyperplane is formulated as: minimize (l/2)llwll A 2 subject to y_i(w • x_i + b) > 1, for all i where 'y_i' represents the class labels.
  • Mean The average value of the EEG signal in the window. Useful for understanding the central tendency.
  • Standard Deviation Measures the amount of variation in the EEG signal. High STD can indicate high neural activity.
  • Skewness Describes the asymmetry of the EEG signal distribution. It helps in understanding the signal's distribution pattern.
  • Kurtosis Measures the 'tailedness' of the EEG signal distribution. High kurtosis indicates infrequent extreme deviations.
  • Root Mean Square Provides a measure of the magnitude of the EEG signal. It is useful for quantifying signal strength.
  • Zero Crossing Rate The rate at which the signal changes sign. It is indicative of the signal's frequency content, where a higher rate can suggest higher frequencies.
  • Delta Band Power (0.5-4 Hz): Associated with deep sleep and brainstem functions.
  • Beta Band Power (13-30 Hz): Associated with active thinking and focus.
  • Peak Frequency The frequency at which the power spectrum has its maximum value.
  • Peak Power Spectral Density The power of the EEG signal at the peak frequency.
  • TFR Time-Frequency Representation
  • TFR Measures the variability of power across frequencies and time, indicating the stability of the signal's frequency content.
  • Sample Entropy Quantifies the complexity and irregularity of the EEG signal. High entropy indicates a more complex signal pattern.
  • a total of 16 features are calculated for each window, and these features are appended for each sniff to create a feature-rich matrix according to a non-limiting embodiment of the present invention.
  • the resulting dataset contains all sniffs, their features, and labels.
  • An SVM classifier with a Radial Basis Function (RBF) kernel is trained on this data using a 70-30 train-test split.
  • the RBF kernel is particularly suitable for this type of data as it can handle non-linear relationships by mapping the input features into higher-dimensional space, allowing the SVM to find a linear separation in this transformed space.
  • the dataset included 447 no sniff epochs (moving between samples, eating reward) with an average duration of 32.96 seconds, 222 non-target sniff epochs with an average duration of 0.40 seconds, and 220 target sniff epochs with an average duration of 0.43 seconds.
  • Figure 24A PCA variance plot (according to a non-limiting embodiment of the present invention):
  • PCA Principal Component Analysis
  • Figure 24B 3D and 2D plots of PC1,PC2, and PC3 (according to a non-limiting embodiment of the present invention)
  • Figure 24C Box Plot of the First 6 Principle Components (according to a non-limiting embodiment of the present invention)
  • This figure illustrates the distribution of features in target and non-target sniffs using box plots for the first six principal components (PCs) according to a non-limiting embodiment of the present invention.
  • Each box plot represents the spread, central tendency, and outliers of principal component scores. This visualization highlights how features differ between target and non-target sniffs, with noticeable differences in spread and outliers across the six PCs.
  • This figure illustrates the performance of the SVM model using three evaluation metrics: the confusion matrix, train-test scores, and 10-fold cross-validation scores according to a non-limiting embodiment of the present invention.
  • Figure 24D(i) Confusion Matrix: The confusion matrix shows the model's classification performance, with 67 true negatives, 53 true positives, 12 false negatives, and 0 false positives, indicating high accuracy in classifying both classes.
  • Figure 24D(ii) Train and Test Scores The bar plot compares the model's accuracy on the training and test datasets, both achieving approximately 0.88. This suggests that the model generalizes well to unseen data with no significant overfitting or underfitting.
  • Figure 24D(iii) Cross-Validation Scores The line plot displays the accuracy scores across 10-fold cross-validation, highlighting variability in model performance across different subsets of the data. Despite some fluctuation, the overall performance remains consistently high, underscoring the model's robustness.
  • the SVM classifier demonstrated strong performance in distinguishing between classes, as evidenced by the key performance metrics.
  • the model achieved an accuracy of 0.91, with a precision of 0.85 and a recall of 1.00 for class 0.0, and a precision of 1.00 and a recall of 0.82 for class 1.0.
  • the fl-scores were 0.92 and 0.90 for classes 0.0 and 1.0, respectively.
  • the macro and weighted averages for precision, recall, and fl -score were consistently high at 0.92 and 0.91, respectively, indicating balanced performance across both classes.
  • the SVM classifier proved to be an effective tool for this classification task, maintaining high precision, recall, and fl -scores across both training and validation phases.
  • the resultant breed-specific FEM serves as the basis for a transfer function that maps the Beagle-optimized EEG interpretation algorithms to the target breed.
  • This transfer function incorporates scaling coefficients for signal amplitude modulation, phase adjustment parameters to account for latency variations, and spatial filters that compensate for breed-specific differences in electrode positioning relative to underlying neuroanatomical structures.
  • the adapted algorithm's efficacy is validated through comparative analysis with breed-specific neurophysiological benchmarks, iteratively refined to achieve optimal performance metrics.
  • Odor Classification Success Entails Ability to Classify Other Modalities of Stimuli according to a non-limiting aspect of the present invention:
  • the adaptation process encompasses: (1) Reconfiguration of data acquisition protocols to capture modality-specific neural responses, including adjustment of sampling rates and frequency bands of interest; (2) Realignment of EEG channel locations to optimally capture the relevant sensory cortices (e.g., occipital lobe for visual stimuli, temporal lobe for auditory stimuli); (3) Refinement of feature extraction algorithms to isolate pertinent signal components, such as event-related potentials (ERPs) or spectral power changes in relevant frequency bands; (4) Modification of the training dataset to incorporate labeled examples of the target stimuli (e.g., images for visual classification, sound snippets for auditory classification); and (5) Fine-tuning of the machine learning model architecture to account for the dimensionality and temporal dynamics specific to each sensory modality.
  • This adaptive framework ensures that the success achieved in odor classification can be effectively translated to the classification of visual and auditory stimuli, maintaining high accuracy and generalizability across diverse sensory inputs.
  • the canine Brain-Computer Interface (BCI) system employs a versatile and extensible multimodal sensor fusion approach to augment traditional EEG data, significantly enhancing the precision, robustness, and interpretability of canine neurophysiological analysis across various experimental paradigms and real-world applications. While the core EEG data, typically acquired from strategically positioned channels at high-fidelity sampling rates, serves as the primary neurophysiological measure, the system's architecture facilitates seamless integration with a diverse array of complementary sensor modalities. This flexibility allows for tailored configurations to meet specific research or operational requirements. For instance, high-frequency infrared (IR) sensors can be employed for precise event epoching, providing temporal markers for stimulus presentation or specific behaviors.
  • IR infrared
  • IMUs Inertial Measurement Units
  • EEG electrocardiography
  • EMG electromyography
  • LSL Lab Streaming Layer
  • sensor fusion allows for immediate cross- modal noise cancellation and artifact removal, enhancing the overall signal quality.
  • Feature-level fusion enables the extraction of complex, multi-modal features that may not be apparent in any single data stream.
  • fusion techniques can combine decisions from multiple modality-specific classifiers, potentially improving the overall accuracy and robustness of cognitive state or behavior predictions.
  • This multi-level fusion approach combined with sophisticated algorithms for event detection and feature extraction, provides a rich contextual framework for interpretation. It enables more nuanced and reliable classification of cognitive states, emotional responses, and sensory processing across a wide spectrum of canine behaviors and environments.
  • the system's modular and extensible design ensures adaptability to emerging sensor technologies and analytical methods, future-proofing the platform for ongoing advancements in the field of canine neuroscience and BCI applications.
  • new sensing modalities and data analysis techniques emerge, they can be readily incorporated into the existing framework, expanding the system's capabilities and opening new avenues for understanding canine cognition and behavior.

Landscapes

  • Engineering & Computer Science (AREA)
  • General Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Dermatology (AREA)
  • General Health & Medical Sciences (AREA)
  • Neurology (AREA)
  • Neurosurgery (AREA)
  • Measuring And Recording Apparatus For Diagnosis (AREA)
  • Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)

Abstract

La présente invention concerne un casque d'interface neuronale directe (IND) personnalisé pour les chiens, un système de collecte et d'analyse de signal d'IND personnalisé, et des procédés d'utilisation de celui-ci pour acquérir et stocker des signaux d'EEG collectés par les capteurs des casques d'IND afin de calculer un état neurologique ou comportemental défini du chien, qui s'appliquent à la détection par le chien d'un danger, le danger pouvant être une matière explosive ou un obstacle physique ; ou à la détection par le chien d'un état de santé ou d'une maladie chez un sujet humain, et dans certains modes de réalisation, l'état neurologique ou comportemental est en corrélation avec la détection de la présence d'une infection, du diabète, du cancer, de l'hypoglycémie, de la narcolepsie, de la maladie de Parkinson ou de crises d'épilepsie.
PCT/US2024/039970 2023-07-27 2024-07-29 Procédé et système d'enregistrement et de décodage de paramètres de l'activité cérébrale de chiens Pending WO2025024839A2 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US202363515855P 2023-07-27 2023-07-27
US63/515,855 2023-07-27

Publications (2)

Publication Number Publication Date
WO2025024839A2 true WO2025024839A2 (fr) 2025-01-30
WO2025024839A3 WO2025024839A3 (fr) 2025-06-19

Family

ID=94375574

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2024/039970 Pending WO2025024839A2 (fr) 2023-07-27 2024-07-29 Procédé et système d'enregistrement et de décodage de paramètres de l'activité cérébrale de chiens

Country Status (1)

Country Link
WO (1) WO2025024839A2 (fr)

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8738139B2 (en) * 2007-08-01 2014-05-27 Bruce Lanning Wireless system for epilepsy monitoring and measurement
US20090112278A1 (en) * 2007-10-30 2009-04-30 Neuropace, Inc. Systems, Methods and Devices for a Skull/Brain Interface
CN115702792A (zh) * 2021-08-17 2023-02-17 陳信彰 犬只脑波情绪识别系统

Also Published As

Publication number Publication date
WO2025024839A3 (fr) 2025-06-19

Similar Documents

Publication Publication Date Title
US11842255B2 (en) Systems and methods for labeling large datasets of physiological records based on unsupervised machine learning
Islam et al. Signal artifacts and techniques for artifacts and noise removal
Klug et al. The BeMoBIL Pipeline for automated analyses of multimodal mobile brain and body imaging data
Petrantonakis et al. Emotion recognition from brain signals using hybrid adaptive filtering and higher order crossings analysis
Casson et al. Electroencephalogram
Gevins Analysis of the electromagnetic signals of the human brain: milestones, obstacles, and goals
Villegas et al. Arm-ECG wireless sensor system for wearable long-term surveillance of heart arrhythmias
Li et al. Emotion recognition of subjects with hearing impairment based on fusion of facial expression and EEG topographic map
US20030073917A1 (en) Patient-specific parameter selection for neurological event detection
Navarro-Sune et al. Riemannian geometry applied to detection of respiratory states from EEG signals: the basis for a brain–ventilator interface
Major et al. A survey of brain computer interfaces and their applications
JP2017537686A (ja) 改善された信号分析に基づくスコアリング方法
Pun et al. Brain-computer interaction research at the Computer Vision and Multimedia Laboratory, University of Geneva
US20110034821A1 (en) Increasing the information transfer rate of brain-computer interfaces
Wu et al. Signal processing for brain–computer interfaces: A review and current perspectives
Amin et al. Single-trial extraction of event-related potentials (ERPs) and classification of visual stimuli by ensemble use of discrete wavelet transform with Huffman coding and machine learning techniques
Essa et al. Brain signals analysis based deep learning methods: Recent advances in the study of non-invasive brain signals
Paul et al. EEG based automated detection of six different eye movement conditions for implementation in personal assistive application
WO2025024839A2 (fr) Procédé et système d'enregistrement et de décodage de paramètres de l'activité cérébrale de chiens
Uribe et al. Physiological Signals Fusion Oriented to Diagnosis-A Review
Rabbani et al. Detection of Different Brain Diseases from EEG Signals Using Hidden Markov Model
Bisht et al. Progress and challenges in physiological artifacts’ detection in electroencephalographic readings
Leong et al. Investigating the brain connectivity signatures of inner peace
Candra Emotion recognition using facial expression and electroencephalography features with support vector machine classifier
KR20150128018A (ko) 뇌-컴퓨터 인터페이스 시스템 기반의 인지 뉴로 피드백 방법

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: 24846612

Country of ref document: EP

Kind code of ref document: A2