WO2024191190A1 - Method, program, and apparatus for screening electrocardiogram signals including noise on basis of neural network model - Google Patents

Abstract

A method for screening electrocardiogram signals including noise, according to an embodiment of the present disclosure, comprises the steps of: obtaining a first electrocardiogram signal through a network unit; inputting the first electrocardiogram signal into a pre-trained neural network model and obtaining first feature information corresponding to the first electrocardiogram signal; identifying at least one cluster including a plurality of pieces of second feature information corresponding to a plurality of second electrocardiogram signals without noise; on the basis of the similarity between the first feature information and the at least one cluster, calculating a noise score of the first electrocardiogram signal; and on the basis of the calculated noise score, determining whether the first electrocardiogram signal is an electrocardiogram signal including noise, wherein the pre-trained neural network model may be pre-trained to restore input electrocardiogram signals on the basis of a training data set including the plurality of second electrocardiogram signals.

Description

Method, program and device for screening electrocardiogram signals including noise based on neural network model

The present disclosure relates to a method for screening out electrocardiogram signals including noise. More specifically, the present disclosure relates to a method for determining whether an electrocardiogram signal includes noise by identifying the characteristics of the electrocardiogram signal using a neural network model.

The electrocardiogram signal is a signal generated by the microcurrent occurring in the heart, and is used to identify the patient's heart condition and heart disease, such as the heart rhythm, decreased blood flow to the heart, and structural abnormalities of the heart. However, the electrocardiogram signal contains noise due to various causes, such as baseline fluctuation, power line interference, muscle contraction and relaxation, and noise of the electrocardiogram measuring device itself.

As such, since it is difficult to accurately diagnose a patient with an electrocardiogram signal containing noise, a process is required to determine in advance whether the electrocardiogram signal measured from the patient contains noise or whether the noise contained in the electrocardiogram signal affects the analysis of the electrocardiogram signal. In this regard, conventional techniques have limited themselves to analyzing the type of noise contained in the electrocardiogram signal, or have required a separate human resource, such as an expert, to analyze the noise in the electrocardiogram signal. Therefore, a method for automatically selecting an electrocardiogram signal containing noise is required.

The present disclosure has been made in response to the aforementioned background technology, and the problem to be solved in the present disclosure is to provide a method, program, and device for selecting an electrocardiogram signal including noise based on a neural network model.

However, the problems to be solved in the present disclosure are not limited to the problems mentioned above, and other problems not mentioned can be clearly understood based on the description below.

According to an embodiment of the present disclosure, a method for selecting an electrocardiogram signal including noise, performed by a computing device including at least one processor, comprises the steps of: obtaining a first electrocardiogram signal through a network unit; inputting the first electrocardiogram signal into a pre-learned neural network model to obtain first feature information corresponding to the first electrocardiogram signal; identifying at least one cluster including a plurality of second feature information corresponding to a plurality of second electrocardiogram signals that do not include noise; calculating a noise score of the first electrocardiogram signal based on a similarity between the first feature information and the at least one cluster; and determining whether the first electrocardiogram signal is an electrocardiogram signal including noise based on the calculated noise score, wherein the pre-learned neural network model is pre-learned to restore an input electrocardiogram signal based on a learning data set composed of the plurality of second electrocardiogram signals.

Alternatively, the pre-learned neural network model may include an encoder that generates feature information of an input electrocardiogram signal and a decoder that reconstructs the input electrocardiogram signal based on the generated feature information, and the step of obtaining the first feature information may include a step of inputting the first electrocardiogram signal to the pre-learned neural network model and obtaining first feature information corresponding to the first electrocardiogram signal through the encoder.

Alternatively, the step of identifying the at least one cluster may include the step of inputting the plurality of second electrocardiogram signals into the pre-trained neural network model to obtain a plurality of second feature information corresponding to the plurality of second electrocardiogram signals, and the step of applying a GMM (Gaussian Mixed Model) algorithm to the obtained plurality of second feature information to identify at least one cluster including the plurality of second feature information within a preset feature space.

Alternatively, the step of calculating the noise score may include the step of repeatedly calculating the noise score of the first electrocardiogram signal by changing the number of the at least one cluster identified within the preset feature space, and the step of determining whether the first electrocardiogram signal is an electrocardiogram signal including noise based on an average value of a plurality of noise scores of the first electrocardiogram signal.

Alternatively, the judging step may determine that the larger the noise score, the more noise is included in the first ECG signal.

Alternatively, the determining step may include a step of identifying the first ECG signal as a third level if the noise score is greater than or equal to a first value, a step of identifying the first ECG signal as a second level if the noise score is less than the first value and greater than or equal to a second value, and a step of identifying the first ECG signal as a first level if the noise score is less than the second value, and the method may include a step of performing an analysis of the first ECG signal if the first ECG signal is identified as the first level.

Alternatively, the step of calculating the noise score may include calculating the noise score based on a Mahalanobis distance between the first feature information and the at least one cluster within a preset feature space.

Alternatively, the step of calculating the noise score may calculate the noise score based on the following formula.

Here, x _obtained is the acquired first ECG signal, f(x _obtained ) is the feature information of the acquired first ECG signal, μ _m is the average value of the feature information of multiple second ECG signals included in the m-th cluster, and ∑ _m is the variance value of the feature information of multiple second ECG signals included in the m-th cluster.

According to one embodiment of the present disclosure, a computer program stored in a computer-readable storage medium, wherein the computer program, when executed on one or more processors, performs operations for selecting an electrocardiogram signal including noise, the operations including: a step of obtaining a first electrocardiogram signal through a network unit; a step of inputting the first electrocardiogram signal to a pre-learned neural network model to obtain first feature information corresponding to the first electrocardiogram signal; a step of identifying at least one cluster including a plurality of second feature information corresponding to a plurality of second electrocardiogram signals that do not include noise; a step of calculating a noise score of the first electrocardiogram signal based on a similarity between the first feature information and the at least one cluster; and a step of determining whether the first electrocardiogram signal is an electrocardiogram signal including noise based on the calculated noise score, wherein the pre-learned neural network model may be pre-learned to restore an input electrocardiogram signal based on a learning data set composed of the plurality of second electrocardiogram signals.

According to one embodiment of the present disclosure, a computing device for selecting an electrocardiogram signal including noise comprises: a processor including at least one core; a memory including program codes executable by the processor; and a network unit; wherein the processor obtains a first electrocardiogram signal through the network unit, inputs the first electrocardiogram signal to a pre-learned neural network model to obtain first feature information corresponding to the first electrocardiogram signal, identifies at least one cluster including a plurality of second feature information corresponding to a plurality of second electrocardiogram signals that do not include noise, calculates a noise score of the first electrocardiogram signal based on a similarity between the first feature information and the at least one cluster, and determines whether the first electrocardiogram signal is an electrocardiogram signal including noise based on the calculated noise score, and wherein the pre-learned neural network model is pre-learned to restore an input electrocardiogram signal based on a learning data set composed of the plurality of second electrocardiogram signals.

Alternatively, the pre-learned neural network model includes an encoder that generates feature information of an input electrocardiogram signal and a decoder that reconstructs the input electrocardiogram signal based on the generated feature information, and the processor can input the first electrocardiogram signal to the pre-learned neural network model and obtain first feature information corresponding to the first electrocardiogram signal through the encoder.

Alternatively, the processor may input the plurality of second electrocardiogram signals into the pre-trained neural network model to obtain a plurality of second feature information corresponding to the plurality of second electrocardiogram signals, and apply a GMM (Gaussian Mixed Model) algorithm to the obtained plurality of second feature information to identify at least one cluster including the plurality of second feature information within a preset feature space.

Alternatively, the processor may repeatedly calculate a noise score of the first ECG signal by changing the number of the at least one cluster identified within the preset feature space, and determine whether the first ECG signal is an ECG signal including noise based on an average value of a plurality of noise scores of the first ECG signal.

Alternatively, the processor may determine that the larger the noise score, the more noise is contained in the first ECG signal.

Alternatively, the processor may identify the first ECG signal as a third level if the noise score is greater than or equal to a first value, identify the first ECG signal as a second level if the noise score is less than the first value and greater than or equal to a second value, identify the first ECG signal as a first level if the noise score is less than the second value, and perform analysis of the first ECG signal if the first ECG signal is identified as the first level.

According to a method for screening an electrocardiogram signal including noise according to an embodiment of the present disclosure, a more accurate diagnosis related to a heart disease can be made by automatically screening an electrocardiogram signal including noise, and manpower and time invested for screening an electrocardiogram signal including noise can be reduced.

FIG. 1 is an exemplary diagram of a computing device for selecting an electrocardiogram signal including noise according to one embodiment of the present disclosure.

FIG. 2 is a schematic block diagram of a computing device according to one embodiment of the present disclosure.

FIG. 3 is a flowchart of a method for controlling a computing device according to an embodiment of the present disclosure.

FIG. 4 is an exemplary diagram showing a neural network model according to one embodiment of the present disclosure.

FIG. 5 is an exemplary diagram showing calculating a noise score of an electrocardiogram signal according to one embodiment of the present disclosure.

FIG. 6 is a detailed configuration diagram of a computing device according to an embodiment of the present disclosure.

Throughout the specification of the present disclosure, the same or similar drawing numbers refer to the same or similar components. In addition, in order to clearly describe the present disclosure, drawing numbers of parts in the drawings that are not related to the description of the present disclosure may be omitted.

The term "or" as used herein is intended to mean an inclusive "or" rather than an exclusive "or." That is, unless otherwise specified herein or the context makes clear, "X employs either A or B" should be understood to mean either one of the natural inclusive permutations. For example, unless otherwise specified herein or the context makes clear, "X employs A or B" can be interpreted to mean either X employs A, X employs B, or X employs both A and B.

The term "and/or" as used herein should be understood to refer to and include all possible combinations of one or more of the relevant concepts listed.

The terms "comprises" and/or "comprising" as used herein should be understood to mean the presence of particular features and/or components. However, it should be understood that the terms "comprises" and/or "comprising" do not exclude the presence or addition of one or more other features, other components, and/or combinations thereof.

Unless otherwise specified in this disclosure or unless the context makes it clear that the singular form is intended to be referred to, the singular should generally be construed to include “one or more.”

Hereinafter, embodiments of the present disclosure will be described in detail with reference to the attached drawings so that those skilled in the art can easily implement the present disclosure. The embodiments presented in the present disclosure are provided so that those skilled in the art can utilize or implement the contents of the present disclosure. Accordingly, various modifications to the embodiments of the present disclosure will be apparent to those skilled in the art. That is, the present disclosure may be implemented in various different forms and is not limited to the embodiments below.

The term "Nth (N is a natural number)" used in the present disclosure can be understood as an expression used to mutually distinguish components of the present disclosure according to a predetermined standard such as a functional viewpoint, a structural viewpoint, or convenience of explanation. For example, components performing different functional roles in the present disclosure can be distinguished as a first component or a second component. However, components that are substantially the same within the technical idea of the present disclosure but should be distinguished for convenience of explanation may also be distinguished as a first component or a second component.

The term "acquisition" as used in this disclosure may be understood to mean not only receiving data via a wired or wireless communication network with an external device or system, but also generating data in an on-device form.

Meanwhile, the term "module" or "unit" used in the present disclosure may be understood as a term referring to an independent functional unit that processes computing resources, such as a computer-related entity, firmware, software or a part thereof, hardware or a part thereof, a combination of software and hardware, etc. In this case, the "module" or "unit" may be a unit composed of a single element, or may be a unit expressed as a combination or set of multiple elements. For example, as a narrow concept, a "module" or "unit" may refer to a hardware element of a computing device or a set thereof, an application program that performs a specific function of software, a processing process implemented through software execution, or a set of instructions for program execution, etc. In addition, as a broad concept, a "module" or "unit" may refer to a computing device itself that constitutes a system, or an application that is executed on a computing device, etc. However, the above-described concept is only an example, and the concept of “module” or “part” may be variously defined within a category understandable to those skilled in the art based on the contents of the present disclosure.

The explanation of the terms set forth above is intended to aid in understanding the present disclosure. Therefore, if the terms set forth above are not explicitly stated as matters limiting the contents of the present disclosure, it should be noted that they are not used to limit the technical ideas of the contents of the present disclosure.

The configuration and operation of specific embodiments of the present disclosure are described in detail below with reference to the attached drawings.

FIG. 1 is an exemplary diagram of a computing device (100) for selecting an electrocardiogram signal including noise according to one embodiment of the present disclosure.

Referring to FIG. 1, a computing device (100) (hereinafter, computing device (100)) for selecting an electrocardiogram signal including noise according to an embodiment of the present disclosure selects an electrocardiogram signal (10) including noise. More specifically, the computing device (100) determines whether an acquired electrocardiogram signal (10) is an electrocardiogram signal including noise. Here, the electrocardiogram signal (10) may be directly acquired by the computing device (100) from a patient, or may be acquired by receiving it from a separate external computing device (200) (e.g., an electrocardiogram measuring device). At this time, the computing device (100) may analyze the acquired electrocardiogram signal (10) and generate diagnostic information about the patient depending on whether the acquired electrocardiogram signal (10) includes noise.

In addition, the computing device (100) identifies the amount of noise included in the electrocardiogram signal (10). In particular, the computing device (100) can classify the electrocardiogram signal (10) into a plurality of levels (or grades) (for example, three levels) and identify the electrocardiogram signal (10) according to the amount of noise included in the electrocardiogram signal (10). Through this, the computing device (100) can request the user to take appropriate action for each level of the electrocardiogram signal (10). For example, the computing device (100) can assign a higher level (for example, level 3) to the electrocardiogram signal (10) as the amount of noise included in the electrocardiogram signal (10) increases, and in this case, the computing device (100) can request the user to replace the electrocardiogram measuring device for the electrocardiogram signal (10) of the highest level (i.e., level 3), and can request re-measurement of the electrocardiogram and correction of the user's posture, etc. for the electrocardiogram signal (10) of the middle level (i.e., level 2). And, for the electrocardiogram signal (10) of the lowest level (i.e., level 1), the electrocardiogram signal (10) can be analyzed and diagnostic information about the patient can be generated.

In order to determine whether an electrocardiogram signal (10) contains noise (or to determine the amount of noise contained in an electrocardiogram signal (10)), a computing device (100) according to an embodiment of the present disclosure may utilize a pre-learned electrocardiogram neural network model (20). Specifically, the computing device (100) may automatically determine whether the electrocardiogram signal (10) contains noise by inputting the acquired electrocardiogram signal (10) into the pre-learned electrocardiogram neural network model (20). That is, the computing device (100) determines whether the electrocardiogram signal (10) contains noise on its own without the input of a separate expert or other human resources. As a result, the time and human resources required to analyze the electrocardiogram signal (10) may be reduced.

Hereinafter, embodiments of the present disclosure related thereto will be described in detail with reference to FIGS. 2 to 5. Meanwhile, the description of the present disclosure described above can be equally applied to the embodiments of the present disclosure below.

FIG. 2 is a schematic block diagram of a computing device (100) according to an embodiment of the present disclosure. FIG. 3 is a flowchart of a method for controlling a computing device (100) according to an embodiment of the present disclosure.

Referring to FIG. 2, a computing device (100) according to an embodiment of the present disclosure includes a network unit (130), a memory (120), and one or more processors (110) (hereinafter, “processors”). However, FIG. 1 is only an example, and the computing device (100) may include other configurations for implementing a computing environment. In addition, only some of the configurations disclosed above may be included in the computing device (100).

The computing device (100) according to one embodiment of the present disclosure may be a hardware device or a part of a hardware device that performs comprehensive processing and calculation of data, or may be a software-based computing environment connected to a communication network. For example, the computing device (100) may be a server that performs an intensive data processing function and shares resources, or may be a client that shares resources through interaction with a server. In addition, the computing device (100) may be a cloud system in which a plurality of servers and clients interact to comprehensively process data. Since the above description is only one example related to the type of the computing device (100), the type of the computing device (100) may be configured in various ways within a category that can be understood by those skilled in the art based on the contents of the present disclosure.

The computing device (100) may be a TV, a desktop PC, a laptop, a smartphone, a tablet PC, a server, etc. Meanwhile, the above-described example is only an example for explaining the computing device (100) and is not necessarily limited to the above-described device.

The processor (110) is connected to the network unit (130) and the memory (120) and controls the overall operation of the computing device (100). The processor (110) according to one embodiment of the present disclosure may be understood as a configuration unit including hardware and/or software for performing computing operations. For example, the processor (110) may read a computer program to perform data processing for machine learning. The processor (110) may process computational processes such as processing of input data for machine learning, feature extraction for machine learning, and error calculation based on backpropagation. The processor (110) for performing such data processing may include a central processing unit (CPU), a general purpose graphics processing unit (GPGPU), a tensor processing unit (TPU), an application specific integrated circuit (ASIC), or a field programmable gate array (FPGA). The type of processor (110) described above is only an example, and thus the type of processor (110) may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

The processor (110) may control one or more components included in the computing device (100) by executing one or more instructions stored in the memory (120), control one or more components as a hardware circuit or chip, or control one or more components as a combination of software and hardware.

In addition, the processor (110) may be implemented as a single core processor including one core, or may be implemented as one or more multicore processors including multiple cores (e.g., homogeneous multi-core or heterogeneous multi-core). When the processor (110) is implemented as a multicore processor, each of the multiple cores included in the multicore processor may include an internal memory of the processor, such as a cache memory (120) and an on-chip memory (120), and a common cache shared by the multiple cores may be included in the multicore processor. In addition, each of the multiple cores (or some of the multiple cores) included in the multicore processor may independently read and execute a program command for implementing a method according to an embodiment of the present disclosure, or all (or some) of the multiple cores may be linked to read and execute a program command for implementing a method according to an embodiment of the present disclosure.

The memory (120) according to one embodiment of the present disclosure may be understood as a configuration unit including hardware and/or software for storing and managing data processed in the computing device (100). That is, the memory (120) may store any type of data generated or determined by the processor (110) and any type of data received by the network unit (130). For example, the memory (120) may include at least one type of storage medium among a flash memory type, a hard disk type, a multimedia card micro type, a card type memory, a RAM (random access memory), a SRAM (static random access memory), a ROM (read-only memory), an EEPROM (electrically erasable programmable read-only memory), a PROM (programmable read-only memory), a magnetic memory, a magnetic disk, and an optical disk. In addition, the memory (120) may also include a database system that controls and manages data in a predetermined system. The type of memory (120) described above is only an example, and thus the type of memory (120) can be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

The memory (120) can manage, by structuring and organizing, data required for the processor (110) to perform operations, combinations of data, and program codes executable by the processor (110). For example, the memory (120) can store medical data received through the network unit (130) described below. The memory (120) can store program codes that operate a neural network model to receive medical data as input and perform learning, program codes that operate a neural network model to receive medical data as input and perform inference according to the intended use of the computing device (100), and processed data generated as the program codes are executed.

For example, a pre-learned neural network model (20) and a learning data set used to learn the pre-learned neural network model may be stored in the memory (120). Here, the learning data set may include a plurality of electrocardiogram signals that do not contain noise.

The network unit (130) is a configuration for the computing device (100) to communicate with an external device, and the computing device (100) can transmit and receive various information through the network unit (130). For example, the computing device (100) can receive an electrocardiogram signal (10) of a patient from an external device through the network unit (130). The network unit (130) can perform data transmission and reception using wired and wireless communication systems such as a local area network (LAN), wideband code division multiple access (WCDMA), long term evolution (LTE), wireless broadband internet (WiBro), fifth generation mobile communication (5G), ultra wide-band, ZigBee, radio frequency (RF) communication, wireless LAN, wireless fidelity, near field communication (NFC), or Bluetooth. The above-described communication systems are only examples, and thus, the wired and wireless communication systems for data transmission and reception of the network unit (130) can be applied in various ways other than the above-described examples.

The network unit (130) can receive data required for the processor (110) to perform calculations through wired or wireless communication with any system or any client, etc. In addition, the network unit (130) can transmit data generated through the calculation of the processor (110) through wired or wireless communication with any system or any client, etc. For example, the network unit (130) can receive medical data through communication with a cloud server that performs tasks such as standardization of medical data, a database in a hospital environment, or a computing device, etc. The network unit (130) can transmit output data of the neural network model, and intermediate data, processed data, etc. derived from the calculation process of the processor (110) through communication with the aforementioned database, server, or computing device, etc. FIG. 4 is an exemplary diagram illustrating a neural network model (20) according to an embodiment of the present disclosure.

Referring to FIG. 3, the processor (110) obtains an electrocardiogram signal through the network unit (130) (S310).

Specifically, the processor (110) obtains the patient's electrocardiogram signal (10) from an external computing device (200) connected to the computing device (100) via the network unit (130). Here, the external computing device (200) may be a device (e.g., an electrocardiogram measuring device) that attaches at least one electrode to the patient's body to detect a microcurrent generated from the patient's heart and generates the patient's electrocardiogram signal (10). However, the present invention is not limited thereto, and the computing device (100) may also directly obtain the patient's electrocardiogram signal (10) via a measuring unit (not shown).

Meanwhile, the processor (110) can obtain various bio-signals (e.g., electromyography signals, brain wave signals, etc.) including the patient's bio-information in addition to the electrocardiogram signal (10).

Hereinafter, for the convenience of explanation of the present disclosure, the acquired electrocardiogram signal (10) is referred to as the first electrocardiogram signal (10).

And, the processor (110) inputs the first electrocardiogram signal (10) into the pre-learned neural network model (20) to obtain first feature information corresponding to the first electrocardiogram signal (10) (S320). For example, the processor (110) may obtain the first electrocardiogram signal (10) for a preset time and input the obtained first electrocardiogram signal (10) into the pre-learned neural network model (20) stored in the memory (120). And, the processor (110) may obtain feature information corresponding to the first electrocardiogram signal (10) input as an output value of the pre-learned neural network model (20).

Hereinafter, for the convenience of explanation of the present disclosure, the characteristic information of the first electrocardiogram signal (10) is referred to as first characteristic information.

Meanwhile, according to one embodiment of the present disclosure, the neural network model (20) may be previously trained to restore the input electrocardiogram signal. At this time, the neural network model (20) may be trained based on a learning data set including a plurality of electrocardiogram signals. In particular, the plurality of electrocardiogram signals included in the learning data set may be clean electrocardiogram signals that do not include noise. The processor (110) may input a plurality of electrocardiogram signals that do not include noise included in the learning data set into the neural network model (20), and train the neural network model (20) to restore each of the input electrocardiogram signals that do not include noise.

For example, the neural network model (20) can be implemented as a convolutional auto encoder (CAE). The CAE can be referred to as a convolutional auto encoder, which is a convolutional neural network including an encoder and a decoder.

The encoder of the neural network model (20) may be composed of layers that perform a process (encoding process) in which input data is compressed to form hidden layers among the multiple layers that constitute the CAE, and the decoder may be composed of layers that perform a process of generating output data from hidden layers, which is the opposite process of the encoder.

Specifically, the encoder may be configured as a one-dimensional convolutional neural network. Accordingly, the encoder may include a convolutional layer and a pooling layer, and the decoder may be configured as a transposed one-dimensional convolutional neural network. Accordingly, the decoder may include an uppooling layer and a transposed convolutional layer. In this case, the encoder and the decoder may be connected to each other as a fully connected layer.

Referring to FIG. 4, the neural network model (20) may include an encoder (21) that generates potential feature information of an input first electrocardiogram signal (10) through a learning process based on the above-described learning data set (3) and a decoder (22) that restores the input first electrocardiogram signal (10) based on the potential feature information generated by the encoder (21). More specifically, the encoder (21) encodes the input first electrocardiogram signal (10) into a potential space to extract low-dimensional potential feature information corresponding to the first electrocardiogram signal (10). Then, the decoder (22) restores the input first electrocardiogram signal (10) based on the potential feature information extracted through the encoder (21). Restoring the first electrocardiogram signal (10) may mean generating an electrocardiogram signal that is identical to or similar to the input first electrocardiogram signal (10). To this end, the neural network model (20) can be trained in advance to minimize the error between the input electrocardiogram signal and the generated (or output) electrocardiogram signal.

Meanwhile, potential feature information may include information such as P wave, T wave, QRS wave, RR interval, ST segment, etc. of the input electrocardiogram signal, or may be defined as a combination of P wave, T wave, QRS wave, RR interval, ST segment, etc. and become information learned by the neural network model (20).

Meanwhile, the neural network model (20) can be implemented with various known network models such as GAN (Generative Adversarial Network).

Referring again to FIG. 4, the processor (110) inputs the first electrocardiogram signal (10) into the pre-learned neural network model (20) and can obtain first feature information, which is potential feature information corresponding to the first electrocardiogram signal (10), through the encoder (21). The first feature information can be produced through the encoder (21) in the form of a matrix, coordinates, etc.

FIG. 5 is an exemplary diagram showing calculating a noise score of an electrocardiogram signal according to one embodiment of the present disclosure.

And, the processor (110) identifies at least one cluster including a plurality of feature information corresponding to a plurality of electrocardiogram signals that do not include noise (S330). Specifically, the processor (110) can identify at least one cluster formed by clustering a plurality of feature information corresponding to a plurality of clean electrocardiogram signals that do not include noise.

The processor (110) can input a plurality of electrocardiogram signals (31) included in the learning data set (3) into a pre-learned neural network model (20) to obtain a plurality of feature information corresponding to the plurality of electrocardiogram signals (31) included in the learning data set (3). Hereinafter, for the convenience of explanation of the present disclosure, a plurality of electrocardiogram signals (31) that do not include noise included in the learning data set (3) are referred to as second electrocardiogram signals (30). In addition, feature information of the second electrocardiogram signal (30) is referred to as second feature information.

Specifically, the processor (110) can acquire a plurality of second electrocardiogram signals (30) from a learning data set used to learn the neural network model (20), and then input the acquired plurality of second electrocardiogram signals (30) into the pre-learned neural network model (20). At this time, the encoder (21) of the pre-learned neural network model (20) encodes each second electrocardiogram signal (30) into a latent space, and outputs a plurality of low-dimensional latent feature information (i.e., second feature information) corresponding to each second electrocardiogram signal (30). Accordingly, the processor (110) can acquire a plurality of second feature information corresponding to the plurality of second electrocardiogram signals (30).

And, the processor (110) can identify at least one cluster formed by clustering a plurality of second feature information. In particular, the processor (110) can identify a plurality of second feature information within a preset feature space. And, the processor (110) can identify at least one cluster formed by clustering a plurality of second feature information.

For example, the processor (110) can identify at least one cluster including the plurality of second feature information within a preset feature space by applying a GMM (Gaussian Mixed Model) algorithm to the acquired plurality of second feature information. Specifically, the processor (110) can identify the Gaussian distribution in which each of the plurality of second feature information is included by calculating a probability value that the plurality of second feature information is generated from each Gaussian distribution, and can identify at least one cluster formed by the plurality of second feature information.

At this time, the preset feature space is a multivariate space, which may be a space composed of variables or combinations of variables represented by feature information (i.e., first and second feature information). For example, the variables may be a criterion for extracting feature information of an electrocardiogram signal from an encoder, or a measurement value related to a feature.

Referring to FIG. 5, the processor (110) can identify a plurality of second feature information within a preset three-dimensional feature space. Then, the processor (110) can identify three clusters (41, 42, 43) formed by grouping the identified plurality of second feature information.

And, the processor (110) calculates the noise score of the first electrocardiogram signal (10) based on the similarity between the first feature information and at least one cluster (S340). Specifically, the processor (110) can identify the first feature information within a preset feature space, calculate the distance between the identified first feature information and at least one cluster, and identify the similarity between the first feature information and the at least one cluster. At this time, the processor (110) can calculate the noise score inversely proportional to the similarity. That is, the processor (110) can determine that the lower the similarity, the more noise is included in the first electrocardiogram signal (10).

Referring again to FIG. 5, the processor (110) can identify the distances between the first feature information and the three clusters (41, 42, 43) identified within the preset feature space, respectively, and calculate a noise score for the first electrocardiogram signal (10) based on each of the identified distances. At this time, the processor (110) can identify an average value of a plurality of second feature information included in the cluster, and calculate a noise score for the first electrocardiogram signal (10) based on the distance between the position corresponding to the identified average value and the position of the first feature information.

For example, when there are multiple clusters, the processor (110) may identify distances between the multiple clusters and the first feature information, respectively, and calculate a noise score based on an average value of the identified multiple distances. That is, referring again to FIG. 5, the processor (110) may calculate an average value of distances d1, d2, and d3 between the multiple clusters, the first feature information, and the three clusters (41, 42, 43), respectively, and calculate a noise score of the first electrocardiogram signal (10) based on the calculated average value. For example, the processor (110) may calculate the noise score by applying a weight to the average value. Here, the weight may be applied differently depending on the number of clusters, environmental factors that measured the electrocardiogram signal, etc.

Or, when there are multiple clusters at least, the processor (110) can identify distances between the multiple clusters and the first feature information, respectively, and calculate a noise score based on the minimum distance among the identified multiple distances. That is, referring again to FIG. 5, the processor (110) can calculate a noise score of the first electrocardiogram signal (10) based on the distance d2 between the second cluster (42) having the closest distance to the first feature information and the first feature information.

Meanwhile, according to one embodiment of the present disclosure, the processor (110) may calculate a noise score based on the Mahalanobis distance between the first feature information and at least one cluster within a preset feature space. Specifically, the processor (110) may measure the Mahalanobis distance between the first feature information and at least one cluster within a preset feature space identified as a multivariate space, and calculate a noise score for the first electrocardiogram signal (10) based on the measured Mahalanobis distance.

At this time, the processor (110) can calculate a noise score based on the first equation below.

Here, x _obtained is the acquired first electrocardiogram signal (10), f(x _obtained ) is the feature information of the acquired first electrocardiogram signal (10), μ _m is the average value of the feature information of the plurality of second electrocardiogram signals (30) included in the mth cluster, and ∑ _m is the variance value of the feature information of the plurality of second electrocardiogram signals (30) included in the mth cluster.

Meanwhile, the processor (110) determines whether the first electrocardiogram signal (10) is an electrocardiogram signal including noise based on the calculated noise score (S350). Alternatively, the processor (110) may determine whether the first electrocardiogram signal (10) is an appropriate signal for analyzing the patient's heart condition and heart disease based on the calculated noise score. That is, if the noise score is very small, the processor (110) may determine that even if the first electrocardiogram signal (10) includes noise, the noise does not affect determining the patient's condition, and may perform analysis of the first electrocardiogram signal (10). If the noise score is equal to or greater than a preset value, the processor (110) may determine that the first electrocardiogram signal (10) is an electrocardiogram signal including noise.

In particular, the processor (110) determines that the first electrocardiogram signal (10) contains a lot of noise as the noise score becomes larger, and if the first electrocardiogram signal (10) is determined to contain a lot of noise, the processor (110) can output information requesting the user to re-measure the electrocardiogram signal through an output interface (not shown) or a display (not shown).

For example, if the noise score is greater than or equal to a first value, the processor (110) may identify the first ECG signal (10) as a third level, if the noise score is less than the first value and greater than or equal to a second value, the processor may identify the first ECG signal (10) as a second level, and if the noise score is less than or equal to the second value, the processor may identify the first ECG signal (10) as a first level. That is, if the noise score is greater than or equal to the first value, the processor (110) may determine that the first ECG signal (10) includes a large amount of noise, and may identify the first ECG signal (10) as a third level. In addition, if the noise score is less than the first value and greater than or equal to the second value, the processor (110) may determine that the first ECG signal (10) includes a normal level of noise, and may identify the first ECG signal (10) as a second level. Additionally, if the noise score is less than the second value, the processor (110) determines that the first electrocardiogram signal (10) has little noise or does not contain noise, and thus can identify the first electrocardiogram signal (10) as the first level.

At this time, the processor (110) outputs the level information of the first electrocardiogram signal (10) through a display or output interface, and in the case of level 3, may request the user to replace the electrocardiogram measuring device, and in the case of level 2, may request re-measurement of the electrocardiogram and correction of the user's posture, etc. In addition, in the case of level 2, the processor (110) may analyze the electrocardiogram signal of the first electrocardiogram signal (10) and generate diagnostic information for the patient. The processor (110) may also generate diagnostic information for the patient by using a neural network model (20) learned to generate diagnostic information by analyzing a pre-learned electrocardiogram signal stored in the memory (120).

Meanwhile, according to one embodiment of the present disclosure, the processor (110) may repeatedly calculate a noise score of the first electrocardiogram signal (10) by changing the number of at least one cluster identified within a preset feature space, and determine whether the first electrocardiogram signal (10) is an electrocardiogram signal including noise based on an average value of a plurality of noise scores of the first electrocardiogram signal (10).

Specifically, the processor (110) can change the number of at least one cluster identified within the preset feature space by adjusting the number of Gaussian distributions set in the GMM algorithm. In FIG. 5, three clusters are illustrated as being identified, but the processor (110) can adjust the number of clusters identified in various ways by changing the number of Gaussian distributions. At this time, the processor (110) can calculate the noise score of the first electrocardiogram signal (10) each time the number of clusters is changed and identified.

For example, when the number of initially set Gaussian distributions is three, the processor (110) can identify three clusters in the preset feature space, as illustrated in FIG. 5. At this time, the processor (110) can calculate a noise score (hereinafter, referred to as a first noise score) based on the similarity between the three clusters and the first feature information. Since the above description is equally applicable in this regard, a detailed description will be omitted. In addition, the processor (110) can change the number of Gaussian distributions to five and identify five clusters including a plurality of second feature information within the preset feature space using a GMM algorithm. In addition, the processor (110) can calculate a noise score (hereinafter, referred to as a second noise score) based on the similarity between the five clusters and the first feature information. In addition, the processor (110) can calculate a final noise score of the first ECG signal (10) as an average value of the first and second noise scores. And, the processor (110) can determine whether the first electrocardiogram signal (10) is an electrocardiogram signal including noise or the level of the first electrocardiogram signal (10) based on the calculated final noise score.

Meanwhile, steps S310 to S350 may be further divided into additional steps or combined into fewer steps, depending on the embodiment of the present disclosure. In addition, some steps may be omitted as needed, and the order between steps may be changed.

FIG. 6 is a detailed configuration diagram of a computing device (100) according to one embodiment of the present disclosure.

Referring to FIG. 6, a computing device (100) according to an embodiment of the present disclosure includes one or more processors (110), a memory (120), a network unit (130), a display (140), a user interface (150), a measurement unit (160), and one or more . A detailed description of any configurations illustrated in FIG. 6 that overlap with those illustrated in FIG. 2 will be omitted.

The display (140) can display various images. Here, the images include both still images and moving images. The display (140) can also output an electrocardiogram signal, and can also output a noise score, a noise level, etc. of the electrocardiogram signal. The display (140) can be implemented as various types of displays, such as an LCD (Liquid Crystal Display Panel), an OLED (Organic Light Emitting Diodes), an LCoS (Liquid Crystal on Silicon), a DLP (Digital Light Processing), etc. In addition, the display (140) can also include a driving circuit, a backlight unit, etc., which can be implemented in a form, such as an a-si TFT, an LTPS (low temperature poly silicon) TFT, an OTFT (organic TFT), etc.

Meanwhile, the display (140) may be implemented as a touch screen by being combined with a touch panel, and in this case, the display (140) may perform the function of an input interface that receives a user's touch input as well as an output interface that outputs an image through the touch screen.

The user interface (150) is a configuration used by the computing device (100) to perform interaction with the user, and may include at least one of a touch sensor, a motion sensor, a button, a jog dial, and a switch, but is not limited thereto.

The measuring unit (160) senses the patient's (1) bio-information to obtain the patient's (1) bio-signal. For example, the electrical signal of the patient's (1) heartbeat can be detected through at least one electrode (161) included in the measuring unit (160) to generate the patient's (1) electrocardiogram signal.

The various embodiments of the present disclosure described above can be combined with additional embodiments and can be modified within a range that can be understood by those skilled in the art in light of the detailed description described above. It should be understood that the embodiments of the present disclosure are illustrative in all respects and not restrictive. For example, each component described as a single component may be implemented in a distributed manner, and likewise, components described as distributed may be implemented in a combined manner. Accordingly, all changes or modifications derived from the meaning, scope, and equivalent concept of the claims of the present disclosure should be interpreted as being included in the scope of the present disclosure.

Claims (15)

A method for selecting an electrocardiogram signal including noise, the method being performed by a computing device including at least one processor, A step of acquiring a first electrocardiogram signal through a network section; A step of inputting the first electrocardiogram signal into the learned neural network model to obtain first feature information corresponding to the first electrocardiogram signal; A step of identifying at least one cluster including a plurality of second feature information corresponding to a plurality of second electrocardiogram signals that do not include noise; A step of calculating a noise score of the first electrocardiogram signal based on the similarity between the first feature information and the at least one cluster; and A step of determining whether the first electrocardiogram signal is an electrocardiogram signal including noise based on the generated noise score; The above-mentioned pre-trained neural network model is, A pre-trained method for restoring an input electrocardiogram signal based on a learning dataset consisting of the above plurality of second electrocardiogram signals. method. In the first paragraph, The above-mentioned pre-trained neural network model is, It includes an encoder that generates feature information of an input electrocardiogram signal and a decoder that restores the input electrocardiogram signal based on the generated feature information. The step of obtaining the above first characteristic information is: A step of inputting the first electrocardiogram signal into the learned neural network model and obtaining first feature information corresponding to the first electrocardiogram signal through the encoder, method. In the first paragraph, The step of identifying at least one cluster comprises: A step of inputting the plurality of second electrocardiogram signals into the learned neural network model to obtain a plurality of second feature information corresponding to the plurality of second electrocardiogram signals; and A step of applying a GMM (Gaussian Mixed Model) algorithm to the acquired plurality of second feature information to identify at least one cluster including the plurality of second feature information within a preset feature space; Including, method. In the third paragraph, The step of calculating the above noise score is: A step of repeatedly calculating a noise score of the first electrocardiogram signal by changing the number of at least one cluster identified within the above-described feature space; and A step of determining whether the first ECG signal is an ECG signal including noise based on an average value of a plurality of noise scores of the first ECG signal; Including, method. In the first paragraph, The above judging steps are: The larger the noise score, the more noise is judged to be contained in the first electrocardiogram signal. method. In paragraph 5, The above judging steps are: A step of identifying the first electrocardiogram signal as a third level if the noise score is greater than or equal to the first value, identifying the first electrocardiogram signal as a second level if the noise score is less than the first value and greater than or equal to the second value, and identifying the first electrocardiogram signal as a first level if the noise score is less than the second value; When the first electrocardiogram signal is identified as the first level, a step of performing analysis of the first electrocardiogram signal; Including, method. In Article 9, The step of calculating the above noise score is: A step of calculating the noise score based on the Mahalanobis distance between the first feature information and the at least one cluster within a preset feature space; Including, method. In Article 6, The step of calculating the above noise score is: The noise score is calculated based on the following formula: method.

Here, x _obtained is the acquired first ECG signal, f(x _obtained ) is the feature information of the acquired first ECG signal, μ _m is the average value of the feature information of multiple second ECG signals included in the m-th cluster, and ∑ _m is the variance value of the feature information of multiple second ECG signals included in the m-th cluster. A computer program stored in a computer-readable storage medium, wherein the computer program, when executed on one or more processors, performs operations for selecting an electrocardiogram signal including noise. The above actions are, A step of acquiring a first electrocardiogram signal through a network section; A step of inputting the first electrocardiogram signal into the learned neural network model to obtain first feature information corresponding to the first electrocardiogram signal; A step of identifying at least one cluster including a plurality of second feature information corresponding to a plurality of second electrocardiogram signals that do not include noise; A step of calculating a noise score of the first electrocardiogram signal based on the similarity between the first feature information and the at least one cluster; and A step of determining whether the first electrocardiogram signal is an electrocardiogram signal including noise based on the generated noise score; The above-mentioned pre-trained neural network model is, A pre-trained method for restoring an input electrocardiogram signal based on a learning dataset consisting of the above plurality of second electrocardiogram signals. Computer program. In a computing device for selecting an electrocardiogram signal including noise, A processor comprising at least one core; A memory containing program codes executable by the processor; network unit; The above processor, Obtaining a first electrocardiogram signal through the above network section, inputting the first electrocardiogram signal into a pre-learned neural network model to obtain first feature information corresponding to the first electrocardiogram signal, identifying at least one cluster including a plurality of second feature information corresponding to a plurality of second electrocardiogram signals that do not include noise, calculating a noise score of the first electrocardiogram signal based on the similarity between the first feature information and the at least one cluster, and determining whether the first electrocardiogram signal is an electrocardiogram signal including noise based on the calculated noise score. The above-mentioned pre-trained neural network model is, A pre-trained method for restoring an input electrocardiogram signal based on a learning dataset consisting of the above plurality of second electrocardiogram signals. Device. In Article 10, The above-mentioned pre-trained neural network model is, An encoder for generating feature information of an input electrocardiogram signal and a decoder for restoring the input electrocardiogram signal based on the generated feature information; The above processor, Inputting the first electrocardiogram signal into the learned neural network model, and obtaining first feature information corresponding to the first electrocardiogram signal through the encoder. device. In Article 10, The above processor, Inputting the plurality of second electrocardiogram signals into the pre-learned neural network model to obtain a plurality of second feature information corresponding to the plurality of second electrocardiogram signals, and applying a GMM (Gaussian Mixed Model) algorithm to the obtained plurality of second feature information to identify at least one cluster including the plurality of second feature information within a preset feature space. Device. In Article 12, The above processor, By changing the number of at least one cluster identified within the above preset feature space, the noise score of the first electrocardiogram signal is repeatedly calculated, and based on the average value of a plurality of noise scores of the first electrocardiogram signal, it is determined whether the first electrocardiogram signal is an electrocardiogram signal including noise. device. In Article 10, The above processor, The larger the noise score, the more noise is judged to be contained in the first electrocardiogram signal. Device. In Article 14, The above processor, If the noise score is greater than or equal to the first value, the first electrocardiogram signal is identified as a third level; if the noise score is less than the first value and greater than or equal to the second value, the first electrocardiogram signal is identified as a second level; if the noise score is less than the second value, the first electrocardiogram signal is identified as a first level; and if the first electrocardiogram signal is identified as the first level, analysis of the first electrocardiogram signal is performed. device.

Images