Method, program, and device for quantifying quality of biological signal

Abstract

The embodiment of the invention discloses a method, a program and a device for quantifying the quality of a biological signal. The method may comprise the steps of: acquiring at least one of a first electrocardiogram dataset labeled based on morphological characteristics of an electrocardiogram signal and a second electrocardiogram dataset labeled based on analysis by a domain expert for interpretation of the electrocardiogram signal; and generating a machine learning model for quantifying the quality of the electrocardiogram signal on the basis of at least one of the first electrocardiogram data set and the second electrocardiogram data set.

Description

Technical Field

[0001] The present invention relates to deep learning technology in the medical field, and specifically, to a method and device capable of quantifying by reflecting the interpretability of biological signals into quality analysis. Background Art

[0002] In order to use signal-based data in application fields, it is important to evaluate the quality of the collected signals. Therefore, multiple scales for evaluating signal quality are being actively studied in various application fields such as acoustics, communication, optics, and medicine. The multiple scales for evaluating signal quality studied in the various application fields as described above are collectively referred to as the signal quality index (SQI: signal quality index).

[0003] Most of the previously studied signal quality indexes reflect how much noise or artifacts are generated in the signal itself into the signal quality. However, in the medical field, even if there is a lot of noise or artifacts in the signal, in most cases, it can be used without problems in clinical judgments such as disease diagnosis or prediction. That is, even if the previously studied signal quality index evaluates the quality as poor based on the presence of noise or artifacts, as long as there is enough information required for clinical interpretation, it should be evaluated as data that can be used in the medical field. Therefore, when evaluating the quality of signals in the medical field, it is necessary to reflect whether the signal is clinically interpretable.

[0004] For example, assume that in order to use an electrocardiogram for diagnosing heart disease A, the quality of the electrocardiogram signal is evaluated. Even an electrocardiogram signal that is judged to be of poor quality based on the previous signal quality index may contain all the information required for diagnosing heart disease A. Therefore, it is inappropriate to simply discard the signal as insufficient information for interpreting heart disease A and evaluate it as difficult to use just because it is judged to be of poor signal quality based on the reference signal quality index. That is, the quality of the signal should be evaluated according to the purpose for which the signal is to be used, so it is necessary to develop an optimized signal quality index according to the application purpose in the medical field. Summary of the Invention

[0005] Technical Problem

[0006] The object of the present invention is to provide a method and device for quantifying the signal quality that can reflect the clinical interpretability of biological signals. Moreover, the present invention provides a method and device for generating a reliable machine learning model for the aforementioned quality quantification.

[0007] However, the technical problems to be solved by the present invention are not limited to the aforementioned problems, and other problems not mentioned above can be clearly understood from the following description.

[0008] Technical Solution

[0009] According to an embodiment of the present invention for solving the problems described above, a method for quantifying the quality of a biological signal executed by a computing device is disclosed. The method may include the following steps: obtaining at least one electrocardiogram (ECG) dataset labeled based on the morphological features of an ECG signal and at least one ECG dataset labeled based on the analysis by a domain expert of the interpretation of the ECG signal; and generating, based on at least one of the first ECG dataset and the second ECG dataset, a machine learning model for quantifying the quality of the ECG signal.

[0010] According to an alternative, the first ECG dataset may be labeled as a first category indicating that the ECG signal is interpretable or a second category indicating that the ECG signal is not interpretable based on the noise grasped based on the morphological features of the ECG signal.

[0011] According to an alternative, the second category may correspond to at least one of the following situations: a situation where there is noise with a preset ratio or more that makes it impossible to specify at least one of the start point and the end point of the waveform of the ECG signal exists with a preset ratio or more; or a situation where there is noise with a preset ratio or more that makes it impossible to identify the R peak of the ECG signal exists with a preset ratio or more.

[0012] According to an alternative, the second ECG dataset may be labeled based on the mode of a plurality of analyses by the domain expert on whether the noise existing in the ECG signal affects disease interpretation.

[0013] According to an alternative, one of the first ECG dataset and the second ECG dataset may be divided into a training dataset, a validation dataset, and a test dataset for generating the machine learning model and then used. Moreover, the remaining one of the first ECG dataset and the second ECG dataset may be used as a test dataset for generating the machine learning model.

[0014] According to an alternative, the machine learning model may include: a first model based on a neural network for inferring the interpretability of an ECG signal based on an ECG dataset; and a second model based on regression analysis for inferring the interpretability of an ECG signal based on an ECG dataset.

[0015] According to an alternative, the steps of generating a machine learning model for quantifying the quality of an electrocardiogram signal based on at least one of the first electrocardiogram dataset and the second electrocardiogram dataset may include the following steps: based on the first training dataset included in the first electrocardiogram dataset, enabling multiple candidate models of the first model to learn; based on the first validation dataset included in the first electrocardiogram dataset, validating the performance of the multiple learned candidate models; based on the first test dataset included in the first electrocardiogram dataset and the second test dataset as the second electrocardiogram dataset, evaluating the performance of at least one candidate model selected through the validation; and based on the specification of the candidate model identified through the evaluation, generating the first model.

[0016] According to an alternative, the steps of generating a machine learning model for quantifying the quality of an electrocardiogram signal based on at least one of the first electrocardiogram dataset and the second electrocardiogram dataset may include the following steps: extracting information on a signal quality index (SQI: signal quality index) used to evaluate the noise of an electrocardiogram signal from the first electrocardiogram dataset to generate an index dataset; based on the third training dataset included in the generated index dataset, enabling multiple candidate models of the second model to learn; based on the third validation dataset included in the generated index dataset, validating the performance of the multiple learned candidate models; based on the third test dataset included in the generated index dataset, evaluating the performance of at least one candidate model selected through the validation; and based on the specification of the candidate model identified through the evaluation, generating the second model.

[0017] An embodiment of the present invention for solving the problem as described above discloses a method for quantifying the quality of a biological signal executed by a computing device. The method may include the following steps: acquiring interpretation object data; and inputting the acquired interpretation object data into a machine learning model to calculate a score indicating the interpretability of the interpretation object data. At this time, the machine learning model can be generated based on at least one of the following electrocardiogram datasets, the electrocardiogram datasets being a first electrocardiogram dataset labeled based on the morphological characteristics of an electrocardiogram signal and a second electrocardiogram dataset labeled based on the analysis performed by a domain expert for the interpretation of an electrocardiogram signal.

[0018] According to an alternative, the method may further include the following steps: inferring whether signals included in the acquired interpretation object data are missing; and combining the calculated score with the inferred presence or absence of signals and determining whether the acquired interpretation object data is interpretable data.

[0019] According to an alternative, whether the signals are missing can be inferred based on whether signal values included in the acquired interpretation object data are blank by a preset ratio or whether waveforms of signals included in the acquired interpretation object data are flat shapes.

[0020] According to an alternative, the step of combining the calculated score with the inferred presence or absence of signals and determining whether the acquired interpretation object data is interpretable data may include the following steps: comparing the calculated score with a threshold value and determining whether noise exists in signals included in the acquired interpretation object data; and determining whether the acquired interpretation object data is interpretable data based on at least one of the presence or absence of noise in signals determined based on the calculated score and the inferred presence or absence of signals.

[0021] According to an alternative, the step of comparing the calculated score with a threshold value and determining whether noise exists in signals included in the acquired interpretation object data may further include the following steps: if the score calculated based on a specific lead is above the threshold value, determining that noise exists in the signals of the specific lead.

[0022] According to an alternative, the step of determining whether the acquired interpretation object data is interpretable data based on at least one of the presence or absence of noise in signals determined based on the calculated score and the inferred presence or absence of signals may include the following steps: determining that signals of a specific lead determined to have noise or inferred to have missing signals based on the calculated score are non-interpretable data.

[0023] An embodiment of the present invention for solving the problems described above discloses a computer program stored in a computer-readable storage medium. When the computer program runs on one or more processors, it executes a plurality of actions for quantifying the quality of biological signals. At this time, the plurality of actions may include the following actions: acquiring at least one electrocardiogram data set from a first electrocardiogram data set labeled based on morphological characteristics of electrocardiogram signals and a second electrocardiogram data set labeled based on empirical judgments of domain experts regarding the interpretation of electrocardiogram signals; and generating a machine learning model for quantifying the quality of electrocardiogram signals based on at least one of the acquired first electrocardiogram data set and the acquired second electrocardiogram data set.

[0024] An embodiment of the present invention for solving the problems described above discloses a computing device for quantifying the quality of a biological signal. The device may include: a processor including at least one core; a memory including a plurality of program codes executable by the processor; and a network unit for obtaining at least one electrocardiogram data set from a first electrocardiogram data set labeled based on the morphological characteristics of an electrocardiogram signal and a second electrocardiogram data set labeled based on the empirical judgment of a domain expert on the interpretation of an electrocardiogram signal. At this time, the processor can generate a machine learning model for quantifying the quality of an electrocardiogram signal based on at least one of the obtained first electrocardiogram data set and the obtained second electrocardiogram data set.

[0025] Effects of the Invention

[0026] The object of the present invention is to provide a method and device for quantifying the quality that can reflect the interpretability of a biological signal. Moreover, the object of the present invention is to provide a method and device for generating a reliable machine learning model for the aforementioned quality quantification. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] Figure 1 is a block diagram of a computing device according to an embodiment of the present invention.

[0028] Figure 2 is a block diagram showing the generation process of a machine learning model according to an embodiment of the present invention.

[0029] Figure 3 is a block diagram showing the generation process of a machine learning model according to an alternative embodiment of the present invention.

[0030] Figure 4 is a block diagram showing the process of quantifying the quality of a biological signal by a computing device according to an embodiment of the present invention.

[0031] Figure 5 is a sequence diagram showing the method for generating a machine learning model for quantifying the quality of a biological signal according to an embodiment of the present invention.

[0032] Figure 6 is a sequence diagram showing the method for quantifying the quality of a biological signal according to an embodiment of the present invention.

[0033] Figure 7 is a sequence diagram summarizing the process of quantifying the quality of a biological signal according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0034] The following is a detailed description of the embodiments of the present invention with reference to the accompanying drawings in order to enable those of ordinary skill in the art to which the present invention pertains (hereinafter referred to as "ordinary skilled artisans in the art") to easily implement the present invention. The embodiments disclosed in the present invention are used to enable ordinary skilled artisans in the art to utilize or implement the content of the present invention. Therefore, various modifications of the embodiments of the present invention are clearly apparent to ordinary skilled artisans in the art. That is, the present invention can be implemented in various different forms and is not limited to the following embodiments.

[0035] Throughout the specification of the present invention, the same or similar graphical symbols represent the same or similar elements. Moreover, graphical symbols of parts irrelevant to the description of the present invention may be omitted in the drawings for the purpose of clearly illustrating the present invention.

[0036] The term "or" used in the present invention does not mean an exclusive "or" but an inclusive "or". That is, if the present invention does not specify or its meaning is unclear in the context of the sentence, "X uses A or B" should be understood to mean one of the natural inclusive permutations. For example, when the present invention does not specify or its meaning is unclear in the context of the sentence, "X uses A or B" can be interpreted as one of the cases where X uses A, or X uses B, or X uses both A and B.

[0037] The term "and / or" used in the present invention should be understood to refer to or include all possible combinations of one or more of the related concepts listed.

[0038] The terms "comprising" and / or "having" used in the present invention should be understood to mean the presence of a specific feature and / or element. However, the terms "comprising" and / or "having" should be understood not to exclude the presence or addition of one or more other features, other elements, and / or combinations thereof.

[0039] If the present invention does not specify or it is not clear from the context of the sentence that it indicates a singular form, the singular form should generally be interpreted to include "one or more".

[0040] The term "the Nth (N is a natural number)" used in the present invention can be understood as an expression used to distinguish the elements of the present invention from each other according to a preset criterion from a functional perspective, a structural perspective, or for convenience of description, etc. For example, elements performing different functions in the present invention can be distinguished as the first element or the second element. However, elements that are essentially the same in the technical spirit of the present invention but need to be distinguished for convenience of description can also be distinguished as the first element or the second element.

[0041] As used in the present invention, the term "acquire" can be understood not only as meaning receiving data through a wireless communication network with an external device or system, but also as meaning generating or receiving data in an on-device form.

[0042] On the other hand, the terms "module" or "unit" used in the present invention can be understood as terms referring to an independent functional unit of processing and 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, a "module" or "unit" can be a unit composed of a single element, or a unit represented by a combination or aggregation of multiple elements. For example, as a protocol concept, a "module" or "unit" can refer to a hardware element or an aggregation thereof in a computing device, an application program that executes a specific function of software, a processing procedure implemented by executing software, or a set of instructions for executing a program. Moreover, as a broad concept, a "module" or "unit" can refer to the computing device itself that constitutes a system or an application program executed in the computing device, etc. However, the foregoing concepts are merely illustrative, and the concepts of "module" or "unit" can be defined in various ways within the scope understandable by those of ordinary skill in the art based on the content of the present invention.

[0043] The term "model" used in the present invention can be understood as a system implemented using mathematical concepts and languages to solve a specific problem, a set of software units for solving a specific problem, or an abstract model of a processing procedure for solving a specific problem. For example, a machine learning "model" can refer to an entire system that operates based on a machine learning algorithm. At this time, the machine learning algorithm can include classification algorithms such as naive Bayes bayes and decision tree, regression analysis algorithms such as linear regression and logistic regression, and deep learning algorithms such as convolutional neural network. The types of machine learning algorithms of the present invention are not limited to the foregoing examples, and can be configured in various ways within the scope understandable by those of ordinary skill in the art based on the foregoing examples.

[0044] The description of the foregoing terms is to assist in understanding the present invention. Therefore, unless the foregoing terms are expressly recited as matters limiting the content of the present invention, they are not used in a sense that limits the technical spirit of the content of the present invention.

[0045] Figure 1 It is a block diagram of a computing device according to an embodiment of the present invention.

[0046] A computing device 100 according to an embodiment of the present invention may be a hardware device or a part of a hardware device that performs integrated processing and operation of data, or may be an operating environment based on software connected via a communication network. For example, the computing device 100 may be a server that serves as a main body for performing intensive data processing functions and sharing resources, or may be a client that shares resources through interaction with the server. Moreover, the computing device 100 may also be a cloud system that enables integrated processing of data through the interaction of multiple servers and multiple clients. The foregoing is merely an example related to the type of the computing device 100, and the type of the computing device 100 can be configured in various ways within the scope understandable by those of ordinary skill in the art based on the content of the present invention.

[0047] Please refer to Figure 1 , a computing device 100 according to an embodiment of the present invention may include a processor 110, a memory 120, and a network unit 130. However, Figure 1 it is merely an example, and the computing device 100 may include other elements for implementing the operating environment. Moreover, the computing device 100 may also include only a part of the disclosed multiple elements.

[0048] The processor 110 according to an embodiment of the present invention can be understood as a constituent unit including hardware and / or software for performing operations. For example, the processor 110 can read a computer program and perform data processing for machine learning. The processor 110 can perform operation processes such as preprocessing of input data for machine learning and error calculation based on backpropagation. The processor 110 for performing data processing as described above can 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), etc. The types of the foregoing processor 110 are merely an illustration, and the types of the processor 110 can be variously configured within the scope understandable by those of ordinary skill in the art based on the content of the present invention.

[0049] The processor 110 can generate a machine learning model for quantifying the quality of an electrocardiogram signal based on electrocardiogram data. The processor 110 can generate a machine learning model for quantifying the quality of an electrocardiogram signal using an electrocardiogram data set labeled with a plurality of factors that affect the interpretation result of the electrocardiogram signal as a basis for clinical judgment. At this time, the electrocardiogram data set used to generate the machine learning model can include at least one electrocardiogram data set labeled based on the morphological features of the waveform affecting the interpretation of the electrocardiogram signal and an electrocardiogram data set labeled based on the analysis performed by domain experts for the interpretation of the electrocardiogram signal. Domain experts can be understood as a group or members of the group who perform clinical judgments such as diagnosing specific diseases after analyzing electrocardiogram signals. That is, the processor 110 can generate a machine learning model that provides a quantitative index for whether the electrocardiogram signal has quality available for clinical interpretation, and the machine learning model utilizes a plurality of features that can be confirmed in the waveform of the electrocardiogram signal and a plurality of empirical bases and judgments used when interpreting the electrocardiogram signal.

[0050] The processor 110 can infer the quality of the electrocardiogram data to be interpreted by 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 interpretable. Moreover, the processor 110 can determine whether the electrocardiogram data to be interpreted is clinically interpretable based on the quality of the electrocardiogram data to be interpreted. Specifically, the processor 110 can input the electrocardiogram data to be interpreted into the machine learning model and generate a quantitative index regarding the quality of each lead signal of the electrocardiogram data to be interpreted. Moreover, the processor 110 can analyze the waveform of the electrocardiogram data and determine whether a signal is missing for each lead of the electrocardiogram data to be interpreted. Moreover, the processor 110 can combine the quantitative index generated by the machine learning model and the determination result of whether each lead signal is missing generated by the waveform analysis, and then determine whether the electrocardiogram data to be interpreted is usable for clinical interpretation for each lead.

[0051] The memory 120 of an embodiment of the present invention can be understood as a hardware and / or software component unit that stores and manages the data processed by the computing device 100. That is, the memory 120 can store any form of data generated or determined by the processor 110 and any form of data received by the network unit 130. For example, the memory 120 may include at least one form of storage medium such as a flash memory type, a hard disk type, a multimedia card micro type, a card 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. Moreover, the memory 120 may also include a database system that controls and manages data in a preset system. The types of the memory 120 described above are merely examples, and the types of the memory 120 can be defined in various ways within the scope understandable by those of ordinary skill in the art based on the content of the present invention.

[0052] The memory 120 can structure and organize data, combinations of data, program code executable by the processor 110, etc. required during the operation of the processor 110, and then manage them. For example, the memory 120 can store electrocardiogram data obtained through the network unit 130 described later. The memory 120 can store program code that drives the processor 110 to generate a machine learning model, program code that enables the processor 110 to use the generated machine learning model to infer the quality of electrocardiogram data, and various data calculated when the program code is executed, etc.

[0053] The network unit 130 according to an embodiment of the present invention can be understood as a component unit that transmits and receives data through any form of known wired or wireless communication system. For example, the network unit 130 can use wired or wireless communication systems such as local area network (LAN), wideband code division multiple access (WCDMA), long term evolution (LTE), wireless broadband internet (WiBro), fifth generation mobile communication (5G), ultra-wideband wireless communication, ZigBee, radio frequency (RF) communication, wireless local area network, wireless fidelity, near field communication (NFC), or Bluetooth to transmit and receive data. The foregoing multiple communication systems are merely examples, and various wired or wireless communication systems for the purpose of data transmission and reception of the network unit 130 can also be applied other than the foregoing examples.

[0054] The network unit 130 can receive data required during the operation of the processor 110 through wired or wireless communication with any system, server, or any client, etc. Moreover, the network unit 130 can transmit data generated through the operation of the processor 110 through wired or wireless communication with any system, server, or any client, etc. 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, etc. The network unit 130 can transmit various data generated through the operation of the processor 110 based on electrocardiogram data through wired or wireless communication with an electrocardiogram detection device or a database in a medical environment, etc.

[0055] Figure 2 It is a block diagram showing the generation process of a machine learning model according to an embodiment of the present invention.

[0056] Please refer to Figure 2 In a computing device 100 according to an embodiment of the present invention, a machine learning model 200 for quantifying the quality of an electrocardiogram (ECG) signal can be generated based on a first ECG data set 10 labeled based on morphological features of the ECG signal and a second ECG data set 20 labeled based on an analysis performed by a domain expert on the interpretation of the ECG signal. The computing device 100 learns multiple candidate models based on machine learning using the first ECG data set 10 and the second ECG data set 20, and then adjusts and evaluates the hyperparameters of the multiple candidate models to generate a machine learning model 200 that infers the clinical interpretability of the ECG signal and provides a quantitative index. At this time, the first ECG data set 10 can be a data set labeled based on whether the ECG signal is interpretable, which is analyzed based on noise grasped based on morphological features of the ECG signal. Moreover, the second ECG data set 20 can be a data set labeled based on the analysis result of whether the ECG signal can be used for the interpretation of a specific disease by an expert in the clinical field.

[0057] For example, the first ECG data set 10 can include the following labels, which are distinguished based on the morphological features presented by the waveform of the ECG signal. The labels included in the first ECG data set 10 can include a first category indicating that the ECG signal is interpretable based on the noise grasped based on the morphological features of the ECG signal, and a second category indicating that the ECG signal is not interpretable based on the noise grasped based on the morphological features of the ECG signal. At this time, the first category or the second category can be distinguished based on whether multiple feature points for clinical interpretation of the ECG signal contain noise to a recognizable degree. Specifically, if at least one of the following conditions is met: a case where noise that makes it impossible to specify at least one of the start point and the end point of the waveform of the ECG signal exists at a preset ratio or more, and a case where noise that makes it impossible to identify the R peak of the ECG signal exists at a preset ratio or more, the data including the ECG signal can be labeled as the second category. At this time, the preset ratio can be set by the producer or user of the computing device 100 according to the purpose of quantifying the quality to achieve that purpose. Moreover, if the waveform of the ECG signal is flat, the data including the ECG signal can be labeled as the second category. If it does not belong to any of the above three conditions, the data including the ECG signal can be labeled as the first category. As described above, the ECG data included in the first ECG data set 10 can be labeled based on the following criterion, which is how the noise grasped based on the morphological features of the ECG signal affects the clinical interpretability of the ECG signal.

[0058] The second electrocardiogram dataset 20 may include labels resulting from an analysis by experts on how the noise contained in the electrocardiogram signals affects clinical interpretability. The labels included in the second electrocardiogram dataset 20 may include a third category of signals that are indicated as being usable for interpretation based on the analysis by experts in the clinical field and a fourth category of signals that are indicated as being unusable for interpretation based on the analysis by experts in the clinical field. The third category or the fourth category is distinguished through the empirical analysis and intuitive analysis of domain experts. Therefore, the second electrocardiogram dataset 20 may be labeled based on the mode of multiple analyses by domain experts on whether the noise present in the electrocardiogram signals affects disease interpretation. That is, the electrocardiogram data included in the second electrocardiogram dataset 20 can be labeled as the third category or the fourth category based on the following results in order to improve the reliability of the labels. The results are presented as the mode in big data equivalent to the set of multiple analysis results of domain experts.

[0059] The computing device 100 can comprehensively use multiple datasets labeled in different ways for training, validating, and testing for generating the machine learning model 200 to generate a high-quality model. Please refer to Figure 2 , the computing device 100 can use either the first electrocardiogram dataset 10 or the second electrocardiogram dataset 20 for the training, validation, and testing of the machine learning model 200 in order to generate the machine learning model 200. Moreover, the computing device 100 can use the remaining one of the first electrocardiogram dataset 10 and the second electrocardiogram dataset 20 for the testing of the machine learning model 200. That is, the computing device 100 comprehensively uses the first electrocardiogram dataset 10 and the second electrocardiogram dataset 20 when generating the machine learning model 200, can verify the performance according to multiple criteria, and thus generate a quantitative model optimized for the electrocardiogram signal quality.

[0060] Specifically, the computing device 100 can use the first electrocardiogram dataset 10 after dividing it into a first training dataset 11, a first validation dataset 15, and a first testing dataset 19. Moreover, the computing device 100 can use the second electrocardiogram dataset 20 as a second testing dataset 25 in order to generate the machine learning model 200. That is, the computing device 100 can use a part of the first electrocardiogram dataset 10 for the learning and validation of multiple candidate models for generating the machine learning model 200 and use the remaining part of the first electrocardiogram dataset 10 and the second electrocardiogram dataset 20 when evaluating at least one candidate model selected through learning and validation. Moreover, the computing device 100 can generate the machine learning model 200 based on the evaluation results of at least one candidate model using the remaining part of the first electrocardiogram dataset 10 and the second electrocardiogram dataset 20.

[0061] For example, the computing device 100 can divide the first electrocardiogram dataset 10 in such a way that the ratio of the first training dataset 11: the first validation dataset 15: the first test dataset 19 becomes 8:1:1, and then use it to generate the machine learning model 200. The computing device 100 can make multiple candidate models designed with various parameters learn based on the first training dataset 11 contained in the first electrocardiogram dataset 10 for generating the machine learning model. At this time, the multiple candidate models can be models that receive the input of electrocardiogram data according to each lead and infer the readability of the electrocardiogram signal for each lead contained in the electrocardiogram data. The computing device 100 can perform performance verification on the multiple learned candidate models based on the first validation dataset 15 contained in the first electrocardiogram dataset 10 and select at least one candidate model through the performance verification. The computing device 100 can use the second electrocardiogram dataset 20 and the first test dataset 19 contained in the first electrocardiogram dataset 10 as the second test dataset 25 to evaluate the performance of at least one candidate model selected through the performance verification. Moreover, the computing device 100 can determine a candidate model whose evaluation value is above a preset benchmark as the final candidate model for generating the machine learning model 200. Moreover, the computing device 100 can generate the machine learning model 200 based on the specifications such as the structure and parameters of the final candidate model. At this time, the machine learning model 200 can be a model that receives the input of electrocardiogram data according to each lead and outputs an electrocardiogram score 30 indicating the readability of the signal for each lead contained in the electrocardiogram data. The electrocardiogram score 30 is an index that quantitatively shows the noise evaluation result of the probability value reflecting the readability of the electrocardiogram signal and can be expressed in numbers, symbols, etc. On the other hand, the ratio of the foregoing datasets is merely an illustration, and the present invention is not limited thereto.

[0062] As described above, the computing device 100 can use a part of the first electrocardiogram dataset 10 and the entire second electrocardiogram dataset 20 for the evaluation of the model generated based on a part of the first electrocardiogram dataset 10 in order to reflect the influence of the noise of the signal on the clinical interpretation such as disease diagnosis in the quantification of the signal quality. That is, the computing device 100 can use the entire second electrocardiogram dataset 20 for the evaluation of the model generated based on the first electrocardiogram dataset 10, so as to reflect whether the electrocardiogram signal is clinically interpretable in the quality of the signal inferred by the machine learning model 200. The computing device 100 generates the machine learning model 200 and can quantify the signal quality in such a way that the signal quality can reflect the clinical interpretability of the signal, rather than simply evaluating the signal quality based on the noise or artifacts of the signal.

[0063] Figure 3 It is a block diagram showing the generation process of a machine learning model according to an alternative embodiment of the present invention.

[0064] Please refer to Figure 3 In an alternative embodiment of the present invention, the machine learning model may include: a first neural network-based model 210 for inferring the readability of an electrocardiogram signal based on an electrocardiogram data set; and a second regression analysis-based model 220 for inferring the readability of an electrocardiogram signal based on an electrocardiogram data set. Specifically, according to an alternative embodiment of the present invention, the computing device 100 can perform ensemble learning on the first neural network-based model 210 and the second regression analysis-based model 220 to generate a machine learning model.

[0065] For example, the first model 210 can include a convolutional neural network based on a residual network (ResNet). The first model 210 can be generated through the following process. First, the input of the first training data set 42 can be received, and multiple candidate models of the first model 210 can be learned. At this time, the learning can be performed based on supervised learning using the following labels, which are labels based on the morphological features of the electrocardiogram signal included in the first training data set 42. After the learning of the multiple candidate models of the first model 210 is completed, the multiple candidate models of the first model 210 can receive the input of the first validation data set 43 and be subjected to performance verification. The multiple candidate models of the first model 210 selected through performance verification then receive the input of the first test data set 44 and the second test data set 55 and are subjected to performance evaluation. Moreover, the first model 210 can be generated based on the specifications of the candidate models that perform well in both the first test data set 44 and the second test data set 55 through performance evaluation. At this time, the first model 210 can be a model that receives the input of electrocardiogram data and outputs a first electrocardiogram score 61 indicating the readability of each lead signal included in the electrocardiogram data. The first electrocardiogram score 61 can be a concept corresponding to the electrocardiogram score 30 described above Figure 2 of the electrocardiogram score.

[0066] The second model 220 may include a model based on logistic regression analysis. The second model 220 may be generated through the following process. First, multiple candidate models of the second model 220 can use the exponential data set 41 extracted from the first electrocardiogram data set 40 as input. The exponential data set 41 may be a data set generated after extracting the following information, which is information about signal quality indices (zerocrossSQI, minSQI, maxSQI, powerSQI, q1SQI, q3SQI, sSQI, kSQI, highfreqSQI, baseSQI, pSQI, etc.) that can be used to evaluate the noise of the electrocardiogram signal. That is, multiple candidate models of the second model 220 can receive the input of the third training data set 45 of the exponential data set 41 and then learn, and the third training data set 45 of the exponential data set 41 contains signal characteristics that can be mathematically expressed according to the signal quality indices used to evaluate noise. After the learning of multiple candidate models of the second model 220 is completed, multiple candidate models of the second model 220 can receive the input of the third validation data set 46 and then undergo performance verification. Multiple candidate models of the second model 220 selected through performance verification can receive the input of the third test data set 47 and then undergo performance evaluation. Moreover, the second model 220 can be generated based on the specifications of the candidate models that exhibit good performance in the third test data set 47 through performance evaluation. At this time, the second model 220 may be a model that receives the input of the exponential data extracted from the electrocardiogram data and outputs the second electrocardiogram score 65 indicating the readability of each lead signal contained in the electrocardiogram signal. The second electrocardiogram score 65 may be a concept corresponding to the electrocardiogram score 30 of Figure 2 as described above.

[0067] As described above, in order to minimize the overfitting problem, which is a limitation of the neural network-based first model 210, the computing device 100 uses the regression analysis-based second model 220 together, so that generalization performance can also be expected for external data other than the data set that has been learned and evaluated. Moreover, since it is possible to confirm the judgment process and basis for what decision is made and on what basis when the second model 220 receives the input of features extracted by an explicit mathematical formula, the computing device 100 uses the second model 220 together to eliminate the limitations of the first model 210 where it is difficult to confirm the judgment basis.

[0068] Meanwhile, since the first model 210 based on a neural network cannot know the basis for judgment, its result value cannot be regarded as reliable. The computing device 100 integrates and learns the first model 210 and the second model 220 to generate a machine learning model, thereby preparing at least a safety device for the reliability problem. Generally speaking, although the first model 210 based on a neural network has a high accuracy but a low ability as a basis, the second model 220 based on regression analysis has a low accuracy but a clear basis. Therefore, the computing device 100 can customize the final result value according to the integration learning form of the first model 210 and the second model 220 to generate it according to the purpose required by the user.

[0069] Figure 4 It is a block diagram showing the process of quantifying the quality of a biological signal of a computing device according to an embodiment of the present invention.

[0070] Please refer to Figure 4 , the computing device 100 according to an embodiment of the present invention can input the data 70 to be interpreted into the first model 210 included in the machine learning model, and then calculate the first electrocardiogram score 81 indicating the interpretability for each lead of the data 70 to be interpreted. At this time, the first model 210 can be generated based on at least one electrocardiogram data set among the electrocardiogram data sets marked based on the morphological features of the electrocardiogram signal and the electrocardiogram data sets marked based on the analysis performed by domain experts on the interpretation of the electrocardiogram signal. Moreover, although Figure 4 not shown, the computing device 100 can extract index data containing information about the signal quality index from the data 70 to be interpreted, and the signal quality index is a signal quality index regarding the noise of the electrocardiogram signal. Moreover, the computing device 100 can input the index data extracted from the data 70 to be interpreted into the second model 220 included in the machine learning model, and then calculate the second electrocardiogram score 85 indicating the interpretability for each lead of the data 70 to be interpreted. The second model 220 can be generated based on the index data set extracted from the electrocardiogram data set marked based on the morphological features of the electrocardiogram signal.

[0071] Moreover, the computing device 100 can infer whether the signal contained in the data 70 to be interpreted is missing. The computing device 100 can infer whether the signal is missing for all leads included in the data 70 to be interpreted, and then generate an inference result 89 regarding whether it is missing for each lead. Specifically, it can be inferred whether the signal is missing based on whether the signal value of each lead of the data 70 to be interpreted is blank by a preset ratio or more or whether the waveform of the signal of each lead of the data 70 to be interpreted is a flat shape. At this time, the preset ratio can be a ratio set by the producer or user of the computing device 100 according to the purpose of quantifying the quality to achieve the purpose.

[0072] The computing device 100 can combine the first electrocardiogram score 81, the second electrocardiogram score 85, and the inference result 89 of whether there is an omission, and then determine whether the data 70 to be interpreted is interpretable data. The computing device 100 comprehensively analyzes the output value of the machine learning model and the inference result of whether there is an omission in each lead signal of the data 70 to be interpreted, and then generates a determination result 90 of whether each lead of the data 70 to be interpreted is interpretable. For example, the computing device 100 can compare the first electrocardiogram score 81 with the first threshold value, and determine that there is noise in the signal of the lead above the first threshold value. Moreover, the computing device 100 can compare the second electrocardiogram score 85 with the second threshold value, and determine that there is noise in the signal of the lead above the second threshold value. At this time, the first threshold value and the second threshold value are values preset by the producer or user of the computing device 100 according to the purpose, and they can be the same value or different values. Moreover, the computing device 100 can determine that there is noise in the signal of the lead whose inference result 89 of whether there is an omission in each lead is determined to be an omission, and determine that there is no noise in the signal of the lead determined not to have an omission. If at least one of the determination result based on the first electrocardiogram score 81, the determination result based on the second electrocardiogram score 85, and the determination result of the inference result 89 of whether there is an omission in each lead is determined to have noise in the lead signal, the computing device 100 can determine that the signal of the lead is uninterpretable data. As described above, the computing device 100 can use multiple analysis results that can be used to determine interpretability in a complementary manner, and can provide quantitative information about signal quality with high reliability.

[0073] Figure 5 It is a sequence diagram showing a method for generating a machine learning model for quantifying the quality of a biological signal according to an embodiment of the present invention.

[0074] Please refer to Figure 5, a computing device 100 according to an embodiment of the present invention can obtain at least one electrocardiogram dataset from a first electrocardiogram dataset labeled based on the morphological characteristics of an electrocardiogram signal and a second electrocardiogram dataset labeled based on the analysis performed by a domain expert on the interpretation of the electrocardiogram signal (step S110). At this time, the first electrocardiogram dataset can be labeled as a first category indicating that the electrocardiogram signal is an interpretable signal or a second category indicating that the electrocardiogram signal is an uninterpretable signal according to the noise grasped based on the morphological characteristics of the electrocardiogram signal. Moreover, the second category can correspond to at least one of a case where noise that makes it impossible to specify at least one of the start point and the end point of the waveform of the electrocardiogram signal exists at a preset ratio or more and a case where noise that makes it impossible to identify the R peak of the electrocardiogram signal exists at a preset ratio or more. The second electrocardiogram dataset can be labeled based on the mode of a plurality of analyses performed by a domain expert on whether the noise present in the electrocardiogram signal affects disease interpretation. For example, the computing device 100 can obtain at least one electrocardiogram dataset from the first electrocardiogram dataset and the second electrocardiogram dataset through wired / wireless communication with a client for the labeling purpose of electrocardiogram data. The computing device 100 can also have an input / output unit and be able to directly perform the labeling of electrocardiogram data and generate at least one electrocardiogram dataset from the first electrocardiogram dataset and the second electrocardiogram dataset.

[0075] The computing device 100 can generate a machine learning model for quantifying the quality of an electrocardiogram signal based on at least one electrocardiogram dataset from the first electrocardiogram dataset and the second electrocardiogram dataset obtained through step S110 (step S120). The computing device 100 can use the first electrocardiogram dataset by dividing it into a training dataset, a validation dataset, and a test dataset in order to generate the machine learning model. Moreover, the computing device 100 can use the second electrocardiogram dataset as a test dataset in order to generate the machine learning model. At this time, as a model that outputs the interpretability of an electrocardiogram signal as a quantitative index based on the electrocardiogram dataset of each lead, the machine learning model can include a first model based on a neural network and a second model based on regression analysis.

[0076] For example, the computing device 100 can enable multiple candidate models of the first model to learn based on the first training data set included in the first electrocardiogram data set. The computing device 100 can verify the performance of the learned multiple candidate models based on the first validation 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 as the second electrocardiogram data set. The computing device 100 uses both the first test data set and the second test data set for evaluation to identify the following model, which can exhibit excellent performance for data labeled with different features and benchmarks. At this time, the model with good performance identified by the computing device 100 can be the model with the highest performance evaluation index or the model above a specific benchmark value. The computing device 100 can generate the first model based on the specifications of the candidate model identified through evaluation. At this time, the specifications of the model are information about the parameters used to construct the neural network, and can include the size, depth, width, learning rules, etc. of the kernel.

[0077] The computing device 100 can extract information about the signal quality index used to evaluate the noise of the electrocardiogram signal from the first electrocardiogram data set to generate an index data set. That is, the computing device 100 can generate an index data set based on multiple features showing the signal quality index based on the electrocardiogram signal noise from the first electrocardiogram data set. The computing device 100 can enable multiple candidate models of the second model to learn based on the third training data set included in the generated index data set. The computing device 100 can verify the performance of the learned multiple candidate models based on the third validation 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 the 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 can be the model with the highest performance evaluation index or the model above a specific benchmark value. The computing device 100 can generate the second model based on the specifications of the candidate model identified through evaluation. At this time, the specifications of the model can be information about the parameters of the model used to perform logistic regression analysis.

[0078] Figure 6 It is a sequence diagram showing a method for quantifying the quality of a biological signal according to an embodiment of the present invention.

[0079] Please refer to Figure 6, the computing device 100 according to an embodiment of the present invention can acquire data of an object to be interpreted (step S210). The data of the object to be interpreted can be understood as electrocardiogram data generated for clinical interpretation applications such as diagnosis or prediction of a specific disease. For example, the computing device 100 can receive the data of the object to be interpreted generated by the electrocardiogram detection device through wired or wireless communication with the electrocardiogram detection device.

[0080] The computing device 100 can input the data of the object to be interpreted acquired through step S210 into a machine learning model and calculate a score indicating the interpretability of the data of the object to be interpreted (step S220). At this time, the machine learning model can be a model generated based on an electrocardiogram data set labeled based on the morphological features of electrocardiogram signals and an electrocardiogram data set labeled based on the analysis performed by a domain expert on the interpretation of electrocardiogram signals. For example, the computing device 100 can input the data of the object to be interpreted into a first model based on a neural network included in the machine learning model and calculate a first score indicating noise reflecting interpretability. Moreover, the computing device 100 can input the data of the object to be interpreted into a second model based on regression analysis included in the machine learning model and calculate a second score indicating noise reflecting interpretability. Figure 5 On the other hand, the computing device 100 can infer whether a signal included in the data of the object to be interpreted acquired through step S210 is missing. The inference of whether a signal is missing can be performed in parallel with step S220 of calculating a score through machine learning. For example, based on a specific lead of the data of the object to be interpreted, if more than 50% of the signal values are blank, the computing device 100 can infer that the signal of that lead is missing. Moreover, based on a specific lead of the data of the object to be interpreted, if the waveform of the signal is flat, the computing device 100 can infer that the signal of that lead is missing. The aforementioned value 50 is merely an example, and thus the ratio value used to determine blankness can be changed according to the purpose of use of the computing device 100.

[0081] For another example, based on a specific lead of the data of the object to be interpreted, if the waveform of the signal is flat, the computing device 100 can infer that the signal of that lead is missing. The aforementioned value 50 is merely an example, and thus the ratio value used to determine blankness can be changed according to the purpose of use of the computing device 100.

[0082] The computing device 100 can combine the scores calculated through step S220 and the omission or not of the signals inferred through the foregoing process, and then determine whether the data to be interpreted is interpretable data. The computing device 100 can compare the scores calculated through step S220 with a threshold value to determine whether there is noise in the signals contained in the data to be interpreted. Moreover, the computing device 100 can determine whether the data to be interpreted is interpretable data based on at least one of the presence or absence of noise in the signals judged based on the scores and the omission or not of the signals. For example, the computing device 100 can compare the first scores generated by the first model based on a neural network with a threshold value to determine whether there is noise in each lead signal of the data to be interpreted. The computing device 100 can compare the second scores generated by the second model based on regression analysis with a threshold value to determine whether there is noise in each lead signal of the data to be interpreted. Moreover, the computing device 100 can determine whether there is noise in each lead signal based on the omission or not of each lead signal. The computing device 100 can synthesize the judgment results for the presence or absence of noise and then determine whether clinical interpretation is possible for each lead of the data to be interpreted.

[0083] Figure 7 It is a sequence diagram that organizes the process of quantifying the quality of biological signals in an embodiment of the present invention. Figure 7 Steps S310, S320 and the foregoing Figure 6 The steps correspond, so the description thereof will be omitted below.

[0084] Please refer to Figure 7 , the computing device 100 of an embodiment of the present invention can determine whether the scores calculated after the data to be interpreted is input into the machine learning model are above the threshold value (step S330). Based on a specific lead, if the score is lower than the threshold value, the computing device 100 can determine that there is no noise in the signal of this lead (step S341). On the contrary, based on a specific lead, if the score is above the threshold value, the computing device 100 can determine that there is noise in the signal of this lead (step 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 the threshold value to determine the presence or absence of noise. 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 can be the same or different.

[0085] The computing device 100 can determine whether a signal is missing based on whether each lead signal of the data to be interpreted is blank at a preset ratio or more or whether the waveform of each lead signal is flat (step S350). If the signal of a specific lead of the data to be interpreted is blank at a preset ratio or more or the waveform of the signal of the specific lead is flat, the computing device 100 can determine that the signal of that lead is missing (step S361). Moreover, the computing device 100 can determine that there is noise in the signal of the lead where the signal is missing. On the contrary, if the signal of a specific lead of the data to be interpreted is blank at less than the preset ratio or the waveform of the signal of the specific lead is not flat, the computing device 100 can determine that the signal of that lead is not missing (step S365). Moreover, the computing device 100 can determine that there is no noise in the signal of the lead where the signal is not missing.

[0086] If it is determined through the foregoing process that there is noise in a specific lead, the computing device 100 can determine that the signal of the specific lead cannot be clinically interpreted (step S370). That is, based on the specific lead, if the score is inferred to be above the critical value and it is determined that there is noise (step S345), or if the signal is determined to be missing (step S361), the computing device 100 can determine that the signal of that lead cannot be interpreted (step S370). On the contrary, if it is determined through the foregoing process that there is no noise in a specific lead, the computing device 100 can determine that the signal of the specific lead can be clinically interpreted (step S380). That is, based on the specific lead, if the score is inferred to be below the critical value and it is determined that there is no noise (step S341), and if the signal is determined to be not missing (step S365), the computing device 100 can determine that the signal of that lead can be interpreted (step S380).

[0087] The various embodiments of the present invention described above can be combined with other different embodiments and can be changed within the scope understandable by those of ordinary skill in the art based on the foregoing detailed description. The embodiments of the present invention are illustrative in all respects and should be construed as not restrictive. For example, the elements described in a single integral form can be implemented dispersedly, and similarly, the elements described in a dispersed form can also be implemented in a combined form. Therefore, all modifications and variations derived from the meaning and scope of the claims of the present invention and their equivalent concepts should be construed as belonging to the scope of the present invention.

Claims

1. A method for quantifying the quality of a biological signal, which is executed by a computing device including at least one processor, characterized in that, Including the following steps: Obtaining at least one electrocardiogram dataset from a first electrocardiogram dataset labeled based on the morphological characteristics of electrocardiogram signals and a second electrocardiogram dataset labeled based on the analysis by domain experts of the interpretation of electrocardiogram signals; And Generating a machine learning model for quantifying the quality of electrocardiogram signals based on at least one of the first electrocardiogram dataset and the second electrocardiogram dataset.

2. The method according to claim 1, characterized in that, The first electrocardiogram dataset is labeled as a first category indicating that the electrocardiogram signal is an interpretable signal or a second category indicating that the electrocardiogram signal is an uninterpretable signal according to the noise grasped based on the morphological characteristics of the electrocardiogram signal.

3. The method according to claim 2, wherein The second category corresponds to at least one of the following situations: A situation where there is more than a preset ratio of noise that makes it impossible to specify at least one of the starting point and the ending point of the waveform of the electrocardiogram signal; Or A situation where there is more than a preset ratio of noise that makes it impossible to identify the R peak of the electrocardiogram signal.

4. The method according to claim 1, characterized in that The second electrocardiogram dataset is labeled based on the mode of multiple analyses by the domain experts on whether the noise present in the electrocardiogram signal affects disease interpretation.

5. The method according to claim 1, wherein One of the first electrocardiogram dataset and the second electrocardiogram dataset is divided into a training dataset, a validation dataset, and a test dataset for generating the machine learning model and then used, The remaining one of the first electrocardiogram dataset and the second electrocardiogram dataset is used as a test dataset for generating the machine learning model.

6. The method according to claim 1, wherein The machine learning model includes: A first model based on a neural network, inferring the interpretability of electrocardiogram signals based on the electrocardiogram dataset; and A second model based on regression analysis, inferring the interpretability of electrocardiogram signals based on the electrocardiogram dataset.

7. The method according to claim 6, wherein The step of generating a machine learning model for quantifying the quality of electrocardiogram signals based on at least one of the first electrocardiogram dataset and the second electrocardiogram dataset includes the following steps: Based on the first training dataset included in the first electrocardiogram dataset, enabling multiple candidate models of the first model to learn; Based on the first validation dataset included in the first electrocardiogram dataset, verifying the performance of the learned multiple candidate models; Based on the first test dataset included in the first electrocardiogram dataset and the second test dataset serving as the second electrocardiogram dataset, evaluating the performance of at least one candidate model selected through the verification; And Based on the specifications of the candidate model identified through the evaluation, generating the first model.

8. The method according to claim 6, wherein The step of generating a machine learning model for quantifying the quality of electrocardiogram signals based on at least one of the first electrocardiogram dataset and the second electrocardiogram dataset includes the following steps: Extract information about the signal quality index used to evaluate the noise of the electrocardiogram signal from the first electrocardiogram data set, and generate an index data set; Based on the third training data set contained in the generated index data set, enable multiple candidate models of the second model to learn; Based on the third validation data set contained in the generated index data set, verify the performance of the learned multiple candidate models; Based on the third test data set contained in the generated index data set, evaluate the performance of at least one candidate model selected through the verification; And Based on the specifications of the candidate model identified through the evaluation, generate the second model.

9. A method for quantifying the quality of a biological signal, executed by a computing device including at least one processor, including the following steps: Obtain data to be interpreted; and Input the obtained data to be interpreted into a machine learning model to calculate a score indicating the interpretability of the data to be interpreted, The machine learning model is generated based on at least one of the following electrocardiogram data sets, and the electrocardiogram data sets are the first electrocardiogram data set labeled based on the morphological characteristics of the electrocardiogram signal and the second electrocardiogram data set labeled based on the analysis performed by domain experts on the interpretation of the electrocardiogram signal.

10. The method according to claim 9, wherein It further includes the following steps: Infer whether the signal contained in the obtained data to be interpreted is missing; and Combine the calculated score with the inferred omission of the signal to determine whether the obtained data to be interpreted is interpretable data.

11. The method according to claim 10, characterized in that, Whether the signal is missing is inferred based on whether the signal value contained in the obtained data to be interpreted is blank by more than a preset ratio or whether the waveform of the signal contained in the obtained data to be interpreted is a flat shape.

12. The method according to claim 10, wherein The step of combining the calculated score with the inferred omission of the signal to determine whether the obtained data to be interpreted is interpretable data includes the following steps: Compare the calculated score with a critical value to determine whether there is noise in the signal contained in the obtained data to be interpreted; and Based on at least one of the presence or absence of noise in the signal judged based on the calculated score and the inferred omission of the signal, determine whether the obtained data to be interpreted is interpretable data.

13. The method according to claim 12, wherein The step of comparing the calculated score with a critical value to determine whether there is noise in the signal contained in the obtained data to be interpreted includes the following steps: If the score calculated based on a specific lead is above the critical value, determine that there is noise in the signal of the specific lead.

14. The method according to claim 12, wherein The step of determining whether the obtained data to be interpreted is interpretable data based on at least one of the presence or absence of noise in the signal judged based on the calculated score and the inferred omission of the signal includes the following steps: Judge the signal of the specific lead determined to have noise or inferred to have a missing signal based on the calculated score as un-interpretable data.

15. A computer program, characterized in that Stored in a computer-readable storage medium, the computer program, when running on one or more processors, performs a plurality of actions for quantifying the quality of a biological signal, the plurality of actions including the following actions: Obtaining at least one electrocardiogram dataset from a first electrocardiogram dataset labeled based on morphological characteristics of an electrocardiogram signal and a second electrocardiogram dataset labeled based on an empirical judgment of an expert's interpretation of an electrocardiogram signal; And Generating a machine learning model for quantifying the quality of an electrocardiogram signal based on at least one of the obtained first electrocardiogram dataset and the obtained second electrocardiogram dataset.

16. A computing device for quantifying the quality of a biological signal, characterized in that Comprising: A processor comprising at least one core; A memory containing a plurality of program codes executable by the processor; and A network unit for obtaining at least one electrocardiogram dataset from a first electrocardiogram dataset labeled based on morphological characteristics of an electrocardiogram signal and a second electrocardiogram dataset labeled based on an empirical judgment of an expert's interpretation of an electrocardiogram signal, The processor generates a machine learning model for quantifying the quality of an electrocardiogram signal based on at least one of the obtained first electrocardiogram dataset and the obtained second electrocardiogram dataset.

Images