Method, program, and device for quantifying quality of biosignal

Abstract

The present disclosure provides a method, a device, and a program for quantifying the quality of a bio-signal. The method for quantifying the quality of a bio-signal according to one embodiment of the present disclosure comprises the steps of: acquiring electrocardiogram data; inputting the acquired electrocardiogram data into a machine learning model so as to acquire a score corresponding to whether the electrocardiogram data is readable; and determining whether the electrocardiogram data is finally readable on the basis of the score, wherein the machine learning model is generated on the basis of an electrocardiogram data set labeled on the basis of analysis by domain experts on the reading of electrocardiogram signals.

Description

Method, program and device for quantifying the quality of biosignals

[0001] The present disclosure relates to deep learning technology in the medical field, and more specifically, to a method and device capable of quantifying the readability of biosignals by reflecting it in quality analysis.

[0002] To utilize signal-based data in applications, assessing the quality of the collected signals is crucial. Therefore, metrics for assessing signal quality are being studied in diverse fields, including acoustics, communications, optics, and medicine. These metrics, collectively referred to as the signal quality index (SQI), are being studied across various application fields.

[0003] Most existing signal quality indices reflect the amount of noise or artifacts present in the signal itself. However, in the medical field, there are many cases where a signal with significant noise or artifacts can be used for clinical judgment, such as disease diagnosis or prediction. In other words, even if signal quality is assessed as poor due to noise or artifacts according to existing signal quality indices, if it contains sufficient information necessary for clinical interpretation, it should be evaluated as usable data in the medical field. Therefore, in the medical field, whether a signal is clinically interpretable needs to be reflected in evaluating signal quality.

[0004] For example, assuming that the quality of an ECG signal is evaluated in order to utilize an ECG to diagnose heart disease A, even an ECG signal that is judged to be of poor quality based on the existing signal quality index may contain all the information necessary to diagnose heart disease A. Therefore, it is inappropriate to unconditionally evaluate a signal as difficult to use because it lacks the information necessary to diagnose heart disease A just because the signal quality is poor based on the reference signal quality index. In other words, since the quality of a signal must be evaluated according to the purpose of utilizing the signal, it is necessary to establish a signal quality index that is optimized for the purpose of utilization in the medical field.

[0005] The present disclosure provides a method and device for quantifying signal quality, which can reflect the clinical readability of biosignals. Furthermore, the present disclosure provides a method and device for generating a reliable machine learning model for the aforementioned quality quantification.

[0006] However, the problems to be solved in this disclosure are not limited to the problems mentioned above, and other problems not mentioned can be clearly understood based on the description below.

[0007] According to one embodiment of the present disclosure for realizing the task as described above, a method for quantifying the quality of a biosignal, performed by a computing device including at least one processor, includes the steps of acquiring electrocardiogram data, inputting the acquired electrocardiogram data into a machine learning model to acquire a score corresponding to whether the electrocardiogram data is readable, and determining the final readability of the electrocardiogram data based on the score, wherein the machine learning model is generated based on a labeled electrocardiogram data set based on an analysis by a domain expert on the readability of the electrocardiogram signal.

[0008] Alternatively, the method further includes a step of estimating whether an electrocardiogram signal included in the acquired electrical conductivity data is missing, and a step of determining whether the electrocardiogram data is finally readable by combining the acquired score and the estimated missingness of the electrocardiogram signal.

[0009] Alternatively, whether the signal is missing can be estimated based on whether the electrocardiogram signal value included in the acquired electrocardiogram data has a blank value greater than a predetermined percentage, or whether the waveform of the electrocardiogram signal included in the acquired electrocardiogram data has a flat shape.

[0010] Alternatively, the step of determining whether the electrocardiogram data is finally readable by combining the acquired score and whether the estimated electrocardiogram signal is missing may include the step of comparing the acquired score with a threshold value to determine whether a signal included in the acquired electrocardiogram data is readable, and the step of determining whether the electrocardiogram data is finally readable based on at least one of whether the signal is readable or whether the estimated signal is missing.

[0011] Alternatively, the step of determining whether the noise exists may include a step of determining that the signal of the specific lead is unreadable when the score obtained corresponding to the specific lead is greater than or equal to a threshold value, when the electrocardiogram data is obtained through multiple leads.

[0012] Alternatively, the step of determining whether the final readability is possible may include a step of determining that the electrocardiogram data is finally readable if the number of leads corresponding to signals determined to have noise among the plurality of leads is less than a preset number when the electrocardiogram data is acquired through a plurality of leads.

[0013] Alternatively, the preset number may vary depending on the number of the plurality of leads.

[0014] Alternatively, the method may include a step of dividing the electrocardiogram data to obtain a plurality of detailed electrocardiogram data, the step of obtaining a score may include a step of inputting the obtained plurality of detailed electrocardiogram data into a machine learning model, and obtaining a score corresponding to whether each detailed electrocardiogram data is readable, and the step of determining whether noise exists may include a step of determining that a signal included in the specific detailed electrocardiogram data is unreadable when the score obtained corresponding to the specific detailed electrocardiogram data is greater than or equal to a threshold value.

[0015] Alternatively, the step of determining whether the final readability is possible may include, when the electrocardiogram data is acquired through a single lead, selecting a plurality of first detailed electrocardiogram data, among the plurality of detailed electrocardiogram data, in which the signal is determined to be readable and in which the signal is not missed, and determining the first detailed electrocardiogram data having the highest acquired score among the selected plurality of first detailed electrocardiogram data as the data to be read.

[0016] According to an embodiment of the present disclosure for realizing the task as described above, a computer program stored in a computer-readable storage medium, wherein the computer program, when executed on one or more processors, performs operations for quantifying the quality of a biosignal, the operations including an operation of obtaining electrocardiogram data, an operation of inputting the obtained electrocardiogram data into a machine learning model to obtain a score corresponding to whether the electrocardiogram data is readable, and an operation of determining whether the electrocardiogram data is finally readable based on the score, wherein the machine learning model can be generated based on a labeled electrocardiogram data set based on an analysis by a domain expert on the reading of the electrocardiogram signal.

[0017] In order to achieve the above-described task, according to an embodiment of the present disclosure, a computing device for quantifying the quality of a biosignal comprises a processor including at least one core and a memory including program codes executable by the processor, wherein the processor obtains electrocardiogram data, inputs the obtained electrocardiogram data into a machine learning model to obtain a score corresponding to whether the electrocardiogram data can be read, and determines whether the electrocardiogram data can be finally read based on the score, and the machine learning model can be generated based on a labeled electrocardiogram data set based on an analysis by a domain expert on the readability of the electrocardiogram signal.

[0018] The present disclosure provides a method and device for quantifying quality that can reflect the readability of biosignals. Furthermore, the present disclosure provides a method and device for generating a reliable machine learning model for the aforementioned quality quantification.

[0019] FIG. 1 is a block diagram of a computing device according to one embodiment of the present disclosure.

[0020] FIG. 2 is a block diagram illustrating a process for generating a machine learning model according to one embodiment of the present disclosure.

[0021] FIG. 3 is a block diagram illustrating a process for generating a machine learning model according to an alternative embodiment of the present disclosure.

[0022] FIG. 4 is a block diagram illustrating a process for quantifying the quality of a biosignal of a computing device according to one embodiment of the present disclosure.

[0023] FIG. 5 is a flowchart illustrating a method for generating a machine learning model for quantifying the quality of a biosignal according to one embodiment of the present disclosure.

[0024] FIG. 6 is a flowchart illustrating a method for quantifying the quality of a biosignal according to one embodiment of the present disclosure.

[0025] FIG. 7 is a flowchart outlining a process for quantifying the quality of a biosignal according to one embodiment of the present disclosure.

[0026] Below, embodiments of the present disclosure are 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 this disclosure are provided to enable those skilled in the art to 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 described below.

[0027] Throughout the specification of this disclosure, identical or similar drawing numbers refer to identical or similar components. Furthermore, for the purpose of clearly describing the disclosure, drawing numbers for parts in the drawings that are not relevant to the description of the disclosure may be omitted.

[0028] 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 clear from context, "X employs A or B" should be understood to mean either of the natural inclusive permutations. For example, unless otherwise specified herein or clear from context, "X employs A or B" can be interpreted to mean either X employs A, X employs B, or X employs both A and B.

[0029] The term "and / or" as used herein should be understood to refer to and include all possible combinations of one or more of the related concepts listed.

[0030] The terms "comprises" and / or "comprising" as used herein should be understood to mean the presence of certain 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.

[0031] 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.”

[0032] The term "Nth (N is a natural number)" used in this disclosure can be understood as an expression used to distinguish components of this disclosure from each other based on a predetermined standard such as a functional perspective, a structural perspective, or convenience of explanation. For example, components performing different functional roles in this disclosure can be distinguished as a first component or a second component. However, components that are substantially the same within the technical spirit of this disclosure but must be distinguished for convenience of explanation may also be distinguished as a first component or a second component.

[0033] The term "acquisition" as used in this disclosure may be understood to refer to generating or receiving data in an on-device form, as well as receiving data via a wireless communication network with an external device or system.

[0034] Meanwhile, the term "module" or "unit" used in the present disclosure can 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, or a combination of software and hardware. At this time, the "module" or "unit" may be a unit composed of a single element, or 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 the computing device itself that constitutes the system, or an application running on the computing device, etc. However, since the above-described concept is only an example, the concept of “module” or “part” may be defined in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

[0035] The term "model" used in this disclosure may be understood as a system implemented using mathematical concepts and language to solve a specific problem, a set of software units to solve a specific problem, or an abstract model regarding a processing process to solve a specific problem. For example, a machine learning "model" may refer to the entire system that performs operations based on a machine learning algorithm. In this case, the machine learning algorithm may include classification algorithms such as naive Bayes and decision trees, regression analysis algorithms such as linear regression and logistic regression, and deep learning algorithms such as convolutional neural networks. The types of machine learning algorithms of this disclosure are not limited to the examples described above, and may be configured in various ways within a range understandable to those skilled in the art based on the examples described above.

[0036] The explanation of the above terms is intended to aid understanding of the present disclosure. Therefore, unless explicitly stated as limiting the contents of the present disclosure, it should be noted that the above terms are not intended to limit the technical ideas of the contents of the present disclosure.

[0037] FIG. 1 is a block diagram of a computing device according to one embodiment of the present disclosure.

[0038] A 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 intensive data processing functions 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 that enables multiple servers and clients to interact with each other to comprehensively process data. Since the above description is only one example related to the type of computing device (100), the type of computing device (100) may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

[0039] Referring to FIG. 1, a computing device (100) according to one embodiment of the present disclosure may include a processor (110), a memory (120), and a network unit (130). However, FIG. 1 is merely an example, and the computing device (100) may include other components for implementing a computing environment. In addition, only some of the disclosed components may be included in the computing device (100).

[0040] 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 perform computational processes such as preprocessing of input data for machine learning, error calculation based on backpropagation, etc. 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 the processor (110) described above is only one example, and thus, the type of the 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.

[0041] The processor (110) may generate a machine learning model for quantifying the quality of an electrocardiogram signal based on electrocardiogram data. The processor (110) may generate a machine learning model for quantifying the quality of an electrocardiogram signal using an electrocardiogram data set labeled with factors that affect the interpretation of the electrocardiogram signal, which serves as the basis for clinical judgment. At this time, the electrocardiogram data set used to generate the machine learning model may include at least one of an electrocardiogram set labeled based on morphological features of the waveform that affect the interpretation of the electrocardiogram signal, or an electrocardiogram data set labeled based on a domain expert's analysis of the interpretation of the electrocardiogram signal. The domain expert may be understood as a group or a member of the group that can interpret electrocardiogram signals to make clinical judgments, such as diagnosing a specific disease. That is, the processor (110) can use features identifiable in the waveform of the electrocardiogram signal and empirical grounds and judgments used to interpret the electrocardiogram signal together to create a machine learning model that can provide a quantitative indicator of whether the electrocardiogram signal is a signal of a quality usable for clinical interpretation.

[0042] The processor (110) can estimate the quality of the electrocardiogram data to be read using the machine learning model generated as described above. At this time, the quality of the electrocardiogram data can indicate whether the electrocardiogram data is clinically readable data. In addition, the processor (110) can determine whether the electrocardiogram data to be read is clinically readable data based on the quality of the electrocardiogram data to be read. Specifically, the processor (110) can input the electrocardiogram data to be read into the machine learning model to generate a quantitative index regarding the quality of the signal for each lead of the electrocardiogram data to be read. In addition, the processor (110) can analyze the waveform of the electrocardiogram data to determine whether a signal is missing for each lead of the electrocardiogram data to be read. In addition, the processor (110) can combine the quantitative index generated through the machine learning model and the result of the determination regarding whether a signal is missing for each lead generated through waveform analysis to determine whether the electrocardiogram data to be read is clinically readable data for each lead.

[0043] 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 random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, or 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 one 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.

[0044] The memory (120) can structure and organize and manage data, combinations of data, and program codes executable by the processor (110) required for the processor (110) to perform operations. For example, the memory (120) can store electrocardiogram data acquired through the network unit (130) described below. The memory (120) can store program codes that cause the processor (110) to generate a machine learning model, program codes that cause the processor (110) to estimate the quality of electrocardiogram data using the generated machine learning model, and various data generated as the program codes are executed.

[0045] The network unit (130) according to one embodiment of the present disclosure may be understood as a component that transmits and receives data through any type of known wired or wireless communication system. For example, the network unit (130) may perform data transmission and reception using a wired or wireless communication system 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. Since the above-described communication systems are only examples, the wired and wireless communication system for data transmission and reception of the network unit (130) may be applied in various ways other than the above-described examples.

[0046] The network unit (130) can receive data necessary for the processor (110) to perform calculations through wired or wireless communication with any system, server, or client. In addition, the network unit (130) can transmit data generated through calculations of the processor (110) through wired or wireless communication with any system, server, or client. For example, the network unit (130) can receive an electrocardiogram data set through wired or wireless communication with an electrocardiogram detection device, a database in a medical environment, or the like. The network unit (130) can transmit various data generated through calculations of the processor (110) based on electrocardiogram data through wired or wireless communication with an electrocardiogram detection device, a database in a medical environment, or the like.

[0047] FIG. 2 is a block diagram illustrating a process for generating a machine learning model according to one embodiment of the present disclosure.

[0048] Referring to FIG. 2, a computing device (100) according to an embodiment of the present disclosure may generate a machine learning model (200) for quantifying the quality of an electrocardiogram signal based on a first electrocardiogram data set (10) labeled based on morphological features of the electrocardiogram signal, and a second electrocardiogram data set (20) labeled based on a domain expert's analysis of the interpretation of the electrocardiogram signal. The computing device (100) may generate a machine learning model (200) that estimates the clinical interpretability of the electrocardiogram signal and provides a quantitative index by learning machine learning-based candidate models based on the first electrocardiogram data set (10) and the second electrocardiogram data set (20), and tuning and evaluating hyper parameters of the candidate models. In this case, the first electrocardiogram data set (10) may be a data set labeled based on whether the electrocardiogram signal is readable, which is analyzed based on noise identified based on morphological features of the electrocardiogram signal. In addition, the second electrocardiogram data set (20) may be a labeled data set based on the results of analysis by an expert in the clinical domain on whether the electrocardiogram signal can be used to diagnose a specific disease.

[0049] For example, the first electrocardiogram data set (10) may include labels according to distinguishable noise based on morphological features of the waveform of the electrocardiogram signal. The labels included in the first electrocardiogram data set (10) may include a first class indicating that the electrocardiogram signal is a readable signal based on noise identified based on the morphological features of the electrocardiogram signal, a second class indicating that the electrocardiogram signal is a readable signal but requires remeasurement, and a third class indicating that the electrocardiogram signal is an unreadable signal based on noise identified based on the morphological features of the electrocardiogram signal. In this case, the first class, the second class, or the third class may be distinguished based on whether feature points used for clinical interpretation of the electrocardiogram signal include noise at a level that can be identified.

[0050] Specifically, if there is a noise that prevents at least one of the start point or the end point of the waveform of the electrocardiogram signal from being identified at a predetermined rate or more, or if there is a noise that prevents the R peak of the electrocardiogram signal from being identified at a predetermined rate or more, the data including the electrocardiogram signal may be labeled as a third class. In this case, the predetermined rate may be a rate set by the manufacturer or user of the computing device (100) to achieve the purpose of quantifying the quality. In addition, if the waveform of the electrocardiogram signal has a flat shape, the data including the electrocardiogram signal may be labeled as a third class. On the other hand, if there is a noise that prevents at least one of the start point or the end point of the waveform of the electrocardiogram signal from being identified at a rate less than the predetermined rate, or if the R peak of the electrocardiogram signal is identified but the noise is present at a rate less than the predetermined rate, the data including the electrocardiogram may be labeled as a second class. If none of the conditions described above apply, the data containing the ECG signal may be labeled as the first class.

[0051] In this way, the electrocardiogram data included in the first electrocardiogram data set (10) can be labeled based on how noise identified based on the morphological characteristics of the electrocardiogram signal affects the clinical readability of the electrocardiogram signal.

[0052] The second electrocardiogram data set (20) may include labels based on expert analysis on how noise contained in the electrocardiogram signal affects the possibility of clinical interpretation. The labels contained in the second electrocardiogram data set (20) may include a fourth class indicating that the signal is available for interpretation based on expert analysis in the clinical domain, a fifth class indicating that the electrocardiogram signal is a readable signal but may require remeasurement, and a sixth class indicating that the signal is unavailable for interpretation based on expert analysis in the clinical domain. Since the fourth, fifth, or sixth classes are distinguished through empirical and intuitive analyses by domain experts, the second electrocardiogram data set (20) may be labeled based on the most frequent values ​​of the analyses by domain experts on whether noise contained in the electrocardiogram signal affects the interpretation of a disease. In other words, the electrocardiogram data included in the second electrocardiogram data set (20) can be labeled as the 4th class, 5th class, or 6th class based on the result that appears as the most frequent value in the big data corresponding to the set of analysis results of domain experts in order to increase the reliability of the label.

[0053] Meanwhile, the second electrocardiogram data set (20) can be obtained through secondary labeling following primary labeling of the electrocardiogram signal based on analysis by an expert in the clinical domain or machine reading. Here, the primary labeling can be performed based on information values ​​regarding features of the electrocardiogram signal. For example, the primary labeling may be performed by assigning labels corresponding to each preset feature of the electrocardiogram signal to the electrocardiogram data, such as the presence or absence of noise in the electrocardiogram signal, the presence or absence of baseline fluctuations, the presence or absence of noise, the presence or absence of low data quality, the presence or absence of identifiable feature information in the electrocardiogram signal, the inclusion of noise for more than 50% of the length of the electrocardiogram signal in the electrocardiogram signal corresponding to one or more leads corresponding to electrocardiogram data containing noise or having low data quality, the inclusion of noise for more than 50% of the length of the electrocardiogram signal in the electrocardiogram signal corresponding to six or more leads corresponding to electrocardiogram data containing noise or having low data quality, the possibility of defining the electrocardiogram data, and the presence or absence of loss of the electrocardiogram data. At this time, an expert in the clinical domain may finally assign labels of the fourth class, the fifth class, and the sixth class based on the one or more labels assigned to each electrocardiogram data, thereby obtaining a second electrocardiogram data set (20).

[0054] Meanwhile, if the labels for the same electrocardiogram signal included in the first electrocardiogram data set (10) and the second electrocardiogram data set (20) are different, the label of the electrocardiogram signal included in the first electrocardiogram data set (10) can be changed to correspond to the label of the electrocardiogram signal included in the second electrocardiogram data set (20). For example, if the same electrocardiogram signal included in the first electrocardiogram data set (10) and the second electrocardiogram data set (20) is assigned a second class label in the first electrocardiogram data set (10) and a sixth class label in the second electrocardiogram data set (20), the label of the first electrocardiogram data set (20) can be changed to a third class label corresponding to the sixth class label. This is to prevent a decrease in psychological trust in the results of machine learning when an ECG signal is determined to be readable through machine learning even though it is judged to be impossible for the user to read with the naked eye, and to provide user-centered machine learning.

[0055] The computing device (100) can generate a high-quality model by comprehensively using data sets labeled in different ways for learning, validating, and testing to generate a machine learning model (200). Referring to FIG. 2, the computing device (100) can use either the first electrocardiogram data set (10) or the second electrocardiogram data set (20) for learning, validating, and testing the machine learning model (200) to generate the machine learning model (200). In addition, the computing device (100) can use the other one of the first electrocardiogram data set (10) and the second electrocardiogram data set (20) for testing the machine learning model (200). That is, the computing device (100) can generate a model that is optimized for quantifying the quality of an electrocardiogram signal by comprehensively using the first electrocardiogram data set (10) and the second electrocardiogram data set (20) to generate the machine learning model (200), and thus, the performance of the model can be verified according to various criteria.

[0056] Specifically, the computing device (100) can use the first electrocardiogram data set (10) by dividing it into a first learning data set (11), a first verification data set (15), and a first test data set (19). In addition, the computing device (100) can use the second electrocardiogram data set (20) as a second test data set (25) to generate a machine learning model (200). That is, the computing device (100) can use a part of the first electrocardiogram data set (10) for learning and verifying candidate models for generating the machine learning model (200), and can use the second electrocardiogram data set (20) together with the remaining part of the first electrocardiogram data set (10) for evaluating at least one candidate model selected through learning and verification. And, the computing device (100) can generate a machine learning model (200) based on the evaluation results for at least one candidate model performed using the remaining part of the first electrocardiogram data set (10) and the second electrocardiogram data set (20).

[0057] For example, the computing device (100) can divide the first electrocardiogram data set (10) so that the first learning data set (11): the first verification data set (15): and the first test data set (19) have a ratio of 8:1:1 and use the divided first electrocardiogram data set (10) to generate a machine learning model (200). The computing device (100) can train candidate models designed with various parameters for generating a machine learning model based on the first learning data set (11) included in the first electrocardiogram data set (10). At this time, the candidate models may be models that receive electrocardiogram data for each lead and estimate the readability of the electrocardiogram signal for each lead included in the electrocardiogram data. The computing device (100) can perform performance verification on the trained candidate models based on the first verification data set (15) included in the first electrocardiogram data set (10) and select at least one candidate model through the performance verification. The computing device (100) can use the second electrocardiogram data set (20) as the second test data set (25) together with the first test data set (19) included in the first electrocardiogram data set (10) to evaluate the performance of at least one candidate model selected through performance verification. Then, the computing device (100) can determine a candidate model whose evaluation value is higher than a predetermined standard as a final candidate model for generating a machine learning model (200). Then, the computing device (100) can generate the machine learning model (200) based on specifications such as the structure and parameters of the final candidate model. At this time, the machine learning model (200) may be a model that receives electrocardiogram data for each lead and outputs an electrocardiogram score (30) indicating the readability of signals for each lead included in the electrocardiogram data. The electrocardiogram score (30) is an index that quantitatively represents the result of a noise evaluation that reflects a probability value for the readability of an electrocardiogram signal, and can be expressed as a number, a symbol, etc.Meanwhile, the ratio of the above-described data set is only an example, and the present disclosure is not limited thereto.

[0058] In this way, the computing device (100) can use the entire second electrocardiogram data set (20) together with a portion of the first electrocardiogram data set (10) to evaluate a model generated based on a portion of the first electrocardiogram data set (10) in order to reflect the influence of signal noise on clinical interpretation, such as disease diagnosis, in quantifying the quality of the signal. In other words, the computing device (100) can use the entire second electrocardiogram data set (20) to evaluate a model generated based on the first electrocardiogram data set (10), thereby allowing the quality of the signal estimated by the machine learning model (200) to reflect whether the electrocardiogram signal is clinically readable. Through the generation of this machine learning model (200), the computing device (100) can quantify the quality of the signal so that the quality of the signal can indicate the clinical readability of the signal, rather than simply evaluating the quality of the signal with noise or artifacts in the signal.

[0059] FIG. 3 is a block diagram illustrating a process for generating a machine learning model according to an alternative embodiment of the present disclosure.

[0060] Referring to FIG. 3, a machine learning model according to an alternative embodiment of the present disclosure may include a neural network-based first model (210) for estimating the readability of an electrocardiogram signal based on an electrocardiogram data set, and a regression analysis-based second model (220) for estimating the readability of an electrocardiogram signal based on the electrocardiogram data set. Specifically, according to an alternative embodiment of the present disclosure, a computing device (100) may generate a machine learning model by ensembling the neural network-based first model (210) and the regression analysis-based second model (220).

[0061] For example, the first model (210) may include a convolutional neural network based on ResNet. This first model (210) may be generated through the following process. First, candidate models of the first model (210) may be trained by receiving a first learning data set (42) as input. At this time, training may be performed based on supervised learning using labels based on morphological features of electrocardiogram signals included in the first learning data set (42). When training of the candidate models of the first model (210) is completed, the candidate models of the first model (210) may be input with a first verification data set (43) to verify their performance. The candidate models of the first model (210) selected through the performance verification may be input with a second test data set (55) together with a first test data set (44) to evaluate their performance. And, based on the specifications of the candidate model that has good performance in both the first test data set (44) and the second test data set (55) through performance evaluation, a first model (210) can be generated. At this time, the first model (210) may be a model that receives electrocardiogram data and outputs a first electrocardiogram score (61) indicating the readability of the signal for each lead included in the electrocardiogram data. The first electrocardiogram score (61) may be a concept corresponding to the electrocardiogram score (30) of FIG. 2 described above.

[0062] The second model (220) may include a model based on logistic regression analysis. This second model (220) may be generated through the following process. First, candidate models of the second model (220) may use as input an index data set (41) extracted from the first electrocardiogram data set (40). The index data set (41) may be a data set generated by extracting information about signal quality indices (zerocrossSQI, minSQI, maxSQI, powerSQI, q1SQI, q3SQI, sSQI, kSQI, highfreqSQI, baseSQI, pSQI, etc.) that can be used to evaluate noise in an electrocardiogram signal or evaluate whether an electrocardiogram signal can be read. That is, the candidate models of the second model (220) can perform learning by receiving a third learning data set (45) of the index data set (41) including mathematically clearly expressible signal features according to a signal quality index for evaluating noise or whether an ECG signal can be read. When the learning of the candidate models of the second model (220) is completed, the candidate models of the second model (220) can receive a third verification data set (46) to verify their performance. The candidate models of the second model (220) selected through the performance verification can receive a third test data set (47) to evaluate their performance. Then, the second model (220) can be generated based on the specifications of the candidate models that have good performance in the third test data set (47) through the performance evaluation. At this time, the second model (220) may be a model that receives index data extracted from ECG data and outputs a second ECG score (65) indicating the readability of a signal for each lead included in the ECG signal. The second electrocardiogram score (65) may be a concept corresponding to the electrocardiogram score (30) of FIG. 2 described above.Accordingly, the electrocardiogram score (30) of FIG. 2 may include at least one of the first electrocardiogram score (61) and the second electrocardiogram score (65).

[0063] Meanwhile, the second model (220) may include a plurality of models trained to output a second electrocardiogram score indicating the readability of a lead-specific signal included in an electrocardiogram signal based on different indices depending on the type of information regarding the index included in the index data set (41) used as the third learning data set (45). For example, the second model (220) may include a model trained to output a second electrocardiogram score related to a low voltage of a waveform of an electrocardiogram signal and a model trained to output a second electrocardiogram score related to whether the waveform of the electrocardiogram signal has a flat shape. At this time, the computing device may obtain a plurality of second electrocardiogram scores.

[0064] In this way, the computing device (100) can expect generalized performance for external data other than the data set on which learning and evaluation were performed by using the second model (220) based on regression analysis together to minimize the overfitting problem, which is a limitation of the first model (210) based on a neural network. In addition, since the second model (220) receives features extracted by a clear mathematical formula as input, and the judgment process and basis for making a decision based on a certain criterion can be confirmed, the computing device (100) can overcome the limitation of the first model (210) that it is difficult to confirm the basis for judgment by using the second model (220) together.

[0065] In addition, the first model (210) based on a neural network can be seen as unreliable in its results because the basis for judgment is unknown. The computing device (100) can provide a minimum safety device for reliability issues by generating a machine learning model by ensembling the first model (210) and the second model (220). In addition, the first model (210) based on a neural network is typically seen as having high accuracy but low basis, and the second model (220) based on regression analysis is typically seen as having low accuracy but clear basis. Therefore, the computing device (100) can generate the final result value by customizing it to suit the user's desired purpose according to the ensemble form of the first model (210) and the second model (220).

[0066] FIG. 4 is a block diagram illustrating a process for quantifying the quality of a biosignal of a computing device according to one embodiment of the present disclosure.

[0067] Referring to FIG. 4, a computing device (100) according to an embodiment of the present disclosure may input data to be read (70) into a first model (210) included in a machine learning model, and may produce a first electrocardiogram score (81) indicating the readability of each lead of the data to be read (70). At this time, the first model (210) may be generated based on at least one of an electrocardiogram data set labeled based on morphological features of the electrocardiogram signal, or an electrocardiogram data set labeled based on analysis by a domain expert on the reading of the electrocardiogram signal. In addition, although not represented in FIG. 4, the computing device (100) may extract index data including information on a signal quality index regarding noise of an electrocardiogram signal from the data to be read (70). In addition, the computing device (100) can input index data extracted from the data to be read (70) into a second model (220) included in the machine learning model to produce a second electrocardiogram score (85) indicating the readability of each lead of the data to be read (70). The second model (220) can be generated based on an index data set extracted from an electrocardiogram data set labeled based on morphological features of the electrocardiogram signal.

[0068] And, the computing device (100) can estimate whether a signal included in the data to be read (70) is missing. The computing device (100) can estimate whether a signal is missing for all leads included in the data to be read (70) and generate a result of estimating whether a signal is missing for each lead (89). Specifically, whether a signal is missing can be estimated based on whether a signal value for each lead of the data to be read (70) is blank by a predetermined ratio or more, or whether a waveform of a signal for each lead of the data to be read (70) is flat. At this time, the predetermined ratio may be a ratio set by the manufacturer or user of the computing device (100) to achieve the purpose of quantifying the quality.

[0069] The computing device (100) can determine whether the data to be read (70) is readable data by combining the first electrocardiogram score (81), the second electrocardiogram score (85), and the estimation result of whether there is a missing signal (89). The computing device (100) can comprehensively analyze the output value of the machine learning model and the estimation result of whether there is a missing signal for each lead of the data to be read (70), and generate a determination result (90) of whether the data to be read (70) is readable for each lead. For example, the computing device (100) can compare the first electrocardiogram score (81) with a first threshold value, and determine that a signal of a lead higher than the first threshold value is unreadable. In particular, the lower the first electrocardiogram score (81), the higher the possibility of reading can be determined by the computing device (100). Meanwhile, the inability to read can also be interpreted as the presence of noise in the signal. Accordingly, the computing device (100) can compare the first threshold value and determine that the signal of the lead that is greater than or equal to the first threshold value contains noise. In addition, the computing device (100) can compare the second electrocardiogram score (85) with the second threshold value and determine that the signal of the lead that is greater than or equal to the second threshold value contains noise or is unreadable. At this time, the first threshold value and the second threshold value are values ​​that the manufacturer or user of the computing device (100) has predetermined according to the purpose, and may be the same value or different values. In addition, the computing device (100) can determine that the signal of the lead that is determined to be missing based on the result of estimating whether or not it is missing per lead is unreadable or contains noise in the signal, and the signal of the lead that is determined not to be missing is readable and contains no noise in the signal or contains as little noise as possible to enable the signal to be read.

[0070] If the computing device (100) determines that noise exists in the signal of a lead (or is determined to be unreadable) based on at least one of the judgment results based on the first electrocardiogram score (81), the judgment result based on the second electrocardiogram score (85), or the judgment result based on the estimation result of whether or not a lead is missing (89), the computing device (100) may determine that the signal of the corresponding lead is ultimately unreadable data. In this way, the computing device (100) can provide quantitative information on the quality of a signal with high reliability by complementarily using the analysis results that can be utilized to determine the possibility of reading.

[0071] Meanwhile, according to one embodiment of the present disclosure, the computing device (100) may determine whether or not to perform a final reading based on the first electrocardiogram score (81), the second electrocardiogram score (85), and the lead-by-lead omission estimation result (89) depending on the method of measuring the electrocardiogram signal included in the electrocardiogram data. Accordingly, the computing device (100) may identify the electrocardiogram measurement method by which the electrocardiogram data was acquired before determining whether or not to perform a reading.

[0072] According to one embodiment of the present disclosure, when electrocardiogram data is acquired through a plurality of leads, the computing device (100) may determine whether the number of leads corresponding to signals determined to be unreadable (or determined to contain noise) among the plurality of leads is less than a preset number. Then, if the number of leads corresponding to signals determined to be unreadable is less than a preset number, the computing device (100) may determine that the electrocardiogram signal of the electrocardiogram data acquired through the plurality of leads is finally readable. In other words, if the number of electrocardiogram data (or leads of the electrocardiogram data) determined to be readable among the plurality of electrocardiogram data acquired through the plurality of leads is greater than or equal to a preset number, the computing device (100) may determine that the electrocardiogram signal of the electrocardiogram data acquired through the remaining leads other than the signals determined to be unreadable is finally readable. This may be determined through the first electrocardiogram score (81) as described above. At this time, the computing device (100) may not consider the second electrocardiogram score (85) and the estimation result (89) of whether or not each lead is missing, or may use them as auxiliary information. Meanwhile, the preset number may be set differently depending on the number of multiple leads.

[0073] For example, based on the first electrocardiogram score (81), if there are 8 or more electrocardiogram data pieces that are determined to be free of noise among the plurality of electrocardiogram data pieces acquired through 12 leads, the computing device may determine that the final reading is possible only with the remaining electrocardiogram data excluding the electrocardiogram data pieces determined to be free of noise.

[0074] Meanwhile, according to one embodiment of the present disclosure, the computing device (100) can segment electrocardiogram data to obtain multiple detailed electrocardiogram data. For example, if the length of the electrocardiogram signal is 30 seconds, the computing device (100) can segment the electrocardiogram signal into 10-second units and obtain detailed electrocardiogram data corresponding to the 10-second electrocardiogram signal. The computing device (100) can move the 10-second unit window by 1 second and apply it to the 30-second electrocardiogram signal. Through this, the computing device (100) can obtain 21 detailed electrocardiogram data. In particular, when the electrocardiogram data is acquired through a single lead, the computing device (100) can segment the electrocardiogram data to obtain multiple detailed electrocardiogram data. In addition, the computing device (100) can input the multiple detailed electrocardiogram data into a machine learning model, and obtain a score corresponding to whether each detailed electrocardiogram data can be read. Here, the acquired score includes at least one of the first electrocardiogram score (81) and the second electrocardiogram score (85). At this time, if the acquired score corresponding to the specific detailed electrocardiogram data is equal to or greater than a threshold value, the computing device can determine that the signal included in the specific detailed electrocardiogram data is unreadable. Then, the computing device can identify the detailed electrocardiogram data that is determined to be readable and in which the signal is determined not to be missing, among the plurality of detailed electrocardiogram data, based on the acquired score (the first electrocardiogram score (81) and the second electrocardiogram score (85)) and the estimation result (89) of whether or not each detailed electrocardiogram data is missing. In other words, for the above-described example, the computing device can determine the detailed electrocardiogram data whose first electrocardiogram score (81) is less than the first threshold value and whose second electrocardiogram score (85) is less than the second threshold value, as the data to be read, among the plurality of detailed electrocardiogram data.In particular, when multiple detailed electrocardiogram data are selected, the computing device (100) can determine the detailed electrocardiogram data with the highest acquired score (e.g., first electrocardiogram score) among the multiple selected 1 detailed electrocardiogram data as the data to be read.

[0075] FIG. 5 is a flowchart illustrating a method for generating a machine learning model for quantifying the quality of a biosignal according to one embodiment of the present disclosure.

[0076] Referring to FIG. 5, a computing device (100) according to an embodiment of the present disclosure may acquire at least one of a first electrocardiogram data set labeled based on morphological features of an electrocardiogram signal, or a second electrocardiogram data set labeled based on a domain expert's analysis of the reading of the electrocardiogram signal (S110). At this time, the first electrocardiogram data set may be labeled as a first class indicating that the electrocardiogram signal is a readable signal, or a second class indicating that the electrocardiogram signal is an unreadable signal, depending on noise identified based on the morphological features of the electrocardiogram signal. In addition, the second class may correspond to at least one of a case where a predetermined ratio or more of noise that makes it impossible to specify at least one of a start point or an end point of a waveform of the electrocardiogram signal exists, or a case where a predetermined ratio or more of noise that makes it impossible to identify an R peak of the electrocardiogram signal exists. The second electrocardiogram data set may be labeled based on the most frequent values ​​of domain expert analyses on whether noise present in the electrocardiogram signal affects disease diagnosis. For example, the computing device (100) may obtain at least one of the first electrocardiogram data set or the second electrocardiogram data set through wired or wireless communication with a client for labeling electrocardiogram data. The computing device (100) may be equipped with an input / output unit to directly perform labeling of the electrocardiogram data and may also generate at least one of the first electrocardiogram data set or the second electrocardiogram data set.

[0077] The computing device (100) may generate a machine learning model for quantifying the quality of an electrocardiogram signal based on at least one of the first electrocardiogram data set or the second electrocardiogram data set acquired through step S110 (S120). The computing device (100) may divide the first electrocardiogram data set into a learning data set, a verification data set, and a test data set and use them to generate the machine learning model. In addition, the computing device (100) may use the second electrocardiogram data set as a test data set to generate the machine learning model. At this time, the machine learning model is a model that outputs the readability of the electrocardiogram signal as a quantified index based on the electrocardiogram data set for each lead, and may include a first model based on a neural network and a second model based on regression analysis.

[0078] For example, the computing device (100) can train candidate models of the first model based on the first learning data set included in the first electrocardiogram data set. The computing device (100) can verify the performance of the trained candidate models based on the first verification data set included in the first electrocardiogram data set. The computing device (100) can evaluate the performance of at least one candidate model selected through verification based on the first test data set included in the first electrocardiogram data set and the second test data set, which is the second electrocardiogram data set. By utilizing both the first test data set and the second test data set for evaluation, the computing device (100) can identify a model that has good performance on data labeled with different features and criteria. At this time, the model with good performance identified by the computing device (100) may be a model with the highest performance evaluation index or a model that is above a specific reference value. The computing device (100) can generate the first model based on the specifications of the candidate models identified through evaluation. At this time, the model specifications are information about parameters for configuring the neural network, and may include kernel size, depth, width, learning rate, etc.

[0079] The computing device (100) can extract information about a signal quality index used to evaluate noise of an electrocardiogram signal from a first electrocardiogram data set, and generate an index data set. That is, the computing device (100) can generate an index data set based on features representing a signal quality index according to noise of an electrocardiogram signal in the first electrocardiogram data set. The computing device (100) can train candidate models of the second model based on a third learning data set included in the generated index data set. The computing device (100) can verify the performance of the trained candidate models based on a third verification data set included in the generated index data set. The computing device (100) can evaluate the performance of at least one candidate model selected through verification based on a third test data set included in the generated index data set. The computing device (100) can identify a model with good performance based on the third test data set. At this time, the model with good performance identified by the computing device (100) may be the model with the highest performance evaluation index, or a model exceeding a certain threshold. The computing device (100) may generate a second model based on the specifications of the candidate model identified through the evaluation. At this time, the model specifications may be information regarding model parameters for performing logistic regression analysis.

[0080] FIG. 6 is a flowchart illustrating a method for quantifying the quality of a biosignal according to one embodiment of the present disclosure.

[0081] Referring to FIG. 6, a computing device (100) according to one embodiment of the present disclosure can acquire data to be read (S210). Data to be read can be understood as electrocardiogram data generated for clinical interpretation, such as diagnosis or prediction of a specific disease. For example, the computing device (100) can receive data to be read generated from an electrocardiogram detection device via wired or wireless communication with the electrocardiogram detection device.

[0082] The computing device (100) can input the data to be read acquired through step S210 into a machine learning model to calculate a score indicating the readability of the data to be read (S220). At this time, the machine learning model may be a model generated based on an electrocardiogram data set labeled based on the morphological features of the electrocardiogram signal according to the above-described FIG. 5, and an electrocardiogram data set labeled based on a domain expert's analysis of the reading of the electrocardiogram signal. For example, the computing device (100) can input the data to be read into a first model based on a neural network included in the machine learning model to calculate a first score corresponding to the readability. The first score may be a score indicating noise reflecting the readability. In addition, the computing device (100) can input the data to be read into a second model based on regression analysis included in the machine learning model to calculate a second score corresponding to the readability.

[0083] Meanwhile, the computing device (100) can estimate whether a signal included in the data to be read obtained through step S210 is missing. The estimation of whether a signal is missing can be performed in parallel with step S220, which calculates a score through machine learning. For example, if a signal value is blank for more than 50% based on a specific lead of the data to be read, the computing device (100) can estimate that the signal of the corresponding lead is missing. In addition, if a signal waveform is flat based on a specific lead of the data to be read, the computing device (100) can estimate that the signal of the corresponding lead is missing. The above-described number 50 is only an example, and thus the numerical value of the ratio for determining blankness can be changed according to the intended use of the computing device (100).

[0084] The computing device (100) can determine whether the data to be read is readable data by combining the score calculated through step S220 and whether the signal is missing estimated through the above-described process. The computing device (100) can determine whether the signal included in the data to be read is readable by comparing the score calculated through step S220 with a threshold value. In addition, the computing device (100) can determine whether the data to be read is readable data based on at least one of whether the signal is readable or whether the signal is missing, as determined based on the score.

[0085] For example, the computing device (100) can compare a first score generated through a first model based on a neural network with a threshold value to determine whether the signal for each lead of the data to be read is readable. The computing device (100) can compare a second score generated through a second model based on regression analysis with a threshold value to determine whether the signal for each lead of the data to be read is readable. In addition, the computing device (100) can determine whether the signal for each lead is readable based on whether the signal for each lead is missing. The computing device (100) can synthesize the respective determination results on whether or not the data to be read is readable to determine whether or not clinical reading is possible for each lead of the data to be read.

[0086] FIG. 7 is a flowchart outlining a process for quantifying the quality of a biosignal according to one embodiment of the present disclosure. Steps S310 and S320 of FIG. 7 correspond to the respective steps of FIG. 6 described above, and therefore, their descriptions will be omitted below.

[0087] Referring to FIG. 7, a computing device (100) according to an embodiment of the present disclosure can determine whether a score calculated by a machine learning model for data to be read is greater than or equal to a threshold value (S330). If the score is less than the threshold value based on a specific lead, the computing device (100) can determine that the signal of the corresponding lead is readable, or that there is no noise or that noise exists to a degree that makes it readable (S341). Conversely, if the score is greater than or equal to the threshold value based on a specific lead, the computing device (100) can determine that the signal of the corresponding lead is unreadable or that there is noise (S345). If the machine learning model includes a first model based on a neural network and a second model based on regression analysis, the computing device (100) can individually compare the scores of each model with a threshold value to determine whether it is readable. At this time, the threshold value compared with the score calculated by the first model and the threshold value compared with the score calculated by the second model may be the same or different.

[0088] The computing device (100) can determine whether a signal is missing based on whether the signal of each lead of the data to be read is blank by a predetermined percentage or more or whether the waveform of the signal of each lead is flat (S350). If the signal of a specific lead of the data to be read is blank by a predetermined percentage or more or the waveform of the signal of the specific lead is flat, the computing device (100) can determine that the corresponding lead has a signal missing (S361). In addition, the computing device (100) can determine that the signal of the lead with a signal missing cannot be read or that noise exists in the signal. Conversely, if the signal of a specific lead of the data to be read is blank by less than a predetermined percentage or the waveform of the signal of the specific lead is not flat, the computing device (100) can determine that the corresponding lead does not have a signal missing (S365). In addition, the computing device (100) can determine that the signal of the lead with no signal missing can be read and does not have noise.

[0089] If it is determined that noise exists in a specific lead through the above-described process, the computing device (100) may determine that the signal of the specific lead is not clinically readable (S370). That is, if it is determined that noise exists because the score is estimated to be greater than or equal to a threshold value based on the specific lead (S345) or if it is determined that a signal is missing (S361), the computing device (100) may determine that the signal of the corresponding lead is not readable (S370). Conversely, if it is determined that noise does not exist in a specific lead through the above-described process, the computing device (100) may determine that the signal of the specific lead is clinically readable (S380). That is, if it is determined that noise does not exist because the score is estimated to be less than a threshold value based on the specific lead (S341) or if it is determined that a signal is not missing (S365), the computing device (100) may determine that the signal of the corresponding lead is readable (S380).

[0090] The various embodiments of the present disclosure described above can be combined with additional embodiments and modified within the scope understood by those skilled in the art in light of the detailed description 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 concepts of the claims of the present disclosure should be construed as being included within the scope of the present disclosure.

Claims

1. A method for quantifying the quality of a biosignal, performed by a computing device including at least one processor, Step of acquiring electrocardiogram data; A step of inputting the acquired electrocardiogram data into a machine learning model to obtain a score corresponding to whether the electrocardiogram data can be read; and A step of determining whether the electrocardiogram data can be finally read based on the score; The above machine learning model is generated based on a labeled electrocardiogram data set based on domain expert analysis of the interpretation of the electrocardiogram signal. method.

2. In paragraph 1, A step of estimating whether or not an electrocardiogram signal included in the acquired new electrical conductivity data is missing is further included; The step of determining whether the final readability is as above is: A step of determining whether the electrocardiogram data can be finally read by combining the obtained score and whether the estimated electrocardiogram signal is missing; Including more, method.

3. In paragraph 2, Whether the above signal is missing or not, It is estimated based on whether the electrocardiogram signal value included in the acquired electrocardiogram data has a blank ratio greater than a predetermined ratio, or whether the waveform of the electrocardiogram signal included in the acquired electrocardiogram data has a flat shape. method.

4. In paragraph 2, The step of determining whether the ECG data can be finally read by combining the obtained score and whether the estimated ECG signal is missing is as follows. A step of comparing the acquired score with a threshold value to determine whether the signal included in the acquired electrocardiogram data is readable; and A step of determining whether the final readability of the electrocardiogram data is determined based on at least one of whether the signal is readable or whether the estimated signal is missing; Including, method.

5. In paragraph 4, The step of determining whether the above noise exists is: When the above electrocardiogram data is acquired through multiple leads, a step of determining that the signal of the specific lead is unreadable when the score acquired corresponding to a specific lead is greater than or equal to a threshold value; Including, method.

6. In paragraph 5, The step of determining whether the final readability is as above is: When the above electrocardiogram data is acquired through multiple leads, if the number of leads corresponding to signals determined to have noise among the multiple leads is less than a preset number, a step of determining that the electrocardiogram data is finally readable; Including, method 7. In paragraph 6, The above preset number is, Depending on the number of the above multiple leads, method.

8. In paragraph 4, A step of dividing the above electrocardiogram data to obtain multiple detailed electrocardiogram data is included. The steps to obtain the above score are: A step of inputting the acquired multiple detailed electrocardiogram data into a machine learning model, and obtaining a score corresponding to whether each detailed electrocardiogram data can be read; including; The step of determining whether the above noise exists is: A step of determining that a signal included in the specific detailed electrocardiogram data is unreadable when a score obtained in response to the specific detailed electrocardiogram data is greater than or equal to a threshold value; including; method.

9. In paragraph 8, The step of determining whether the final readability is as above is: Including a step of selecting a plurality of first detailed electrocardiogram data, in which the signal is determined to be readable and the signal is not missing, from among the plurality of detailed electrocardiogram data, and determining the first detailed electrocardiogram data having the highest acquired score from among the selected plurality of first detailed electrocardiogram data as the data to be read; method.

10. A computer program stored in a computer-readable storage medium, wherein the computer program, when executed on one or more processors, performs operations for quantifying the quality of a biosignal. The above actions are, The act of acquiring electrocardiogram data; An operation of inputting the acquired electrocardiogram data into a machine learning model to obtain a score corresponding to whether the electrocardiogram data can be read; and An operation for determining whether the electrocardiogram data can be finally read based on the score; The above machine learning model is generated based on a labeled electrocardiogram data set based on domain expert analysis of the interpretation of the electrocardiogram signal. Computer program.

11. A computing device for quantifying the quality of a biosignal, a processor comprising at least one core; and A memory including program codes executable by the processor; The above processor, Obtaining electrocardiogram data, inputting the obtained electrocardiogram data into a machine learning model to obtain a score corresponding to whether the electrocardiogram data can be read, and determining the final readability of the electrocardiogram data based on the score. The above machine learning model is generated based on a labeled electrocardiogram data set based on domain expert analysis of the interpretation of the electrocardiogram signal. Computing device.

Images