Method, program, and device for predicting health condition using electrocardiogram

Abstract

According to one embodiment of the present disclosure, disclosed are a method, program, and apparatus for predicting a health condition using an electrocardiogram performed by a computing device. The method may comprise: a step of acquiring electrocardiogram data and cardiac arrest data including whether or not cardiac arrest occurs in a subject on which the electrocardiogram data is detected and a cardiac arrest occurrence time; a step of extracting valid data from the electrocardiogram data by analyzing a missing value of the electrocardiogram data or the cardiac arrest data; and a step for generating learning data by marking the extracted valid data on the basis of whether or not cardiac arrest has occurred and the time at which cardiac arrest has occurred.

Description

Technical Field

[0001] The present disclosure relates to artificial intelligence technology in the medical field, and specifically, to the following method: learning an artificial intelligence model based on the results of filtering electrocardiogram data reflecting cardiac arrest status, and using the learned model to predict health status. Background Art

[0002] An electrocardiogram is a test that records the electrical activity of the heart. An electrocardiogram is a relatively simple and cost-effective test that can confirm the health of the heart, which plays an important role in the early diagnosis and management of heart disease. For example, the electrocardiogram signal measured by an electrocardiogram can be used to confirm whether the parts of the heart are working properly, whether the size and position of the heart are normal, and whether the heart muscle is damaged. Moreover, the electrocardiogram signal can be used to diagnose a variety of heart-related problems based on these confirmations and can be used to predict a person's health status.

[0003] In addition, with the development of artificial intelligence technology, there are more and more attempts to use artificial intelligence technology to predict health conditions based on electrocardiograms. With the introduction of artificial intelligence technology, the most important part is to ensure the appropriate data required for artificial intelligence learning. Electrocardiograms record heart activity, so cardiac arrest has an important impact on the interpretation of electrocardiograms. Most existing technologies do not reflect the occurrence and status of cardiac arrest in the data ensured for artificial intelligence learning. Summary of the invention

[0004] Technical issues

[0005] The purpose of the present disclosure is to provide a method for learning an artificial intelligence model based on the result of filtering electrocardiogram data reflecting the state of cardiac arrest, and using the learned model to predict health conditions.

[0006] However, the problems to be solved by the present disclosure are not limited to the problems mentioned above, and other problems not mentioned can be clearly understood from the following description.

[0007] Technical Solution

[0008] According to an embodiment of the present disclosure for achieving the above-mentioned purpose, a method for predicting health status using an electrocardiogram performed by a computing device is disclosed. The method may include: obtaining electrocardiogram data and cardiac arrest data including detecting whether a subject of the electrocardiogram data has cardiac arrest and the time of cardiac arrest; analyzing missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data; and labeling the extracted valid data based on whether cardiac arrest has occurred and the time of cardiac arrest to generate learning data.

[0009] Optionally, the step of analyzing missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data may include: the step of analyzing the time of occurrence of cardiac arrest and clinical basis of cardiac arrest recorded in the cardiac arrest data to determine a cutoff time for filtering the electrocardiogram data; and the step of filtering the electrocardiogram data of subjects who have suffered cardiac arrest within the cutoff time from the acquired electrocardiogram data to extract valid data.

[0010] Optionally, the step of analyzing the time of occurrence of cardiac arrest recorded in the cardiac arrest data and the clinical evidence regarding cardiac arrest to determine the cutoff time for filtering the electrocardiogram data may include: determining as the cutoff time one of a first candidate time calculated based on a clinical judgment of changes in the electrocardiogram that occurred before the occurrence of cardiac arrest, a second candidate time that is clinically judged to be a time when measures to prevent cardiac arrest can be taken when cardiac arrest is predicted to occur, or a third candidate time corresponding to the error between the time of occurrence of cardiac arrest recorded in the cardiac arrest data and the actual time of occurrence of cardiac arrest of the object for which the electrocardiogram data was detected.

[0011] Optionally, the deadline may be determined as a maximum value among the first candidate time, the second candidate time, or the third candidate time.

[0012] Optionally, the step of generating learning data by marking the extracted valid data based on whether cardiac arrest occurs and the time when cardiac arrest occurs may include: marking the data of objects in the valid data who have experienced cardiac arrest within a preset time after the cut-off time as a cardiac arrest group, and marking the data of objects in the valid data who have not experienced cardiac arrest or have not died within the preset time after the cut-off time as a normal group.

[0013] Optionally, the step of analyzing the missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data may include: when more than one lead among the N leads of the electrocardiogram data is missing, the step of excluding the electrocardiogram data in which more than one lead among the N leads is missing from the valid data.

[0014] Optionally, the step of analyzing the missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data may include: when there is a lead with missing values ​​for more than a specified time among the N leads of the electrocardiogram data, the step of excluding the electrocardiogram data of the lead with missing values ​​for more than a specified time among the N leads of the electrocardiogram data from the valid data.

[0015] Optionally, the method may further include: a step of inputting the learning data into a deep learning model to learn the deep learning model in such a manner that the deep learning model calculates the possibility of cardiac arrest occurring.

[0016] According to an embodiment of the present disclosure for achieving the aforementioned purpose, a method for predicting health status using an electrocardiogram performed by a computing device is disclosed. The method may include: a step of acquiring first electrocardiogram data; and a step of inputting the first electrocardiogram data into a pre-learned deep learning model to predict the health status of an object detecting the electrocardiogram data. At this time, the deep learning model may learn based on learning data generated by marking valid data extracted from the second electrocardiogram data based on whether cardiac arrest occurs and the time of cardiac arrest. In addition, the valid data may be extracted from the second electrocardiogram data based on the result of analyzing the missing values ​​of the second electrocardiogram data or the cardiac arrest data including whether cardiac arrest occurs and the time of cardiac arrest of the object detecting the second electrocardiogram data.

[0017] Optionally, the prediction result derived by inputting the first electrocardiogram into the pre-learned deep learning model may include a score value representing the health status of the object detecting the electrocardiogram data. And, the closer the score value is to a first threshold, the healthier it is, and the closer it is to a second threshold, the less healthy it is.

[0018] According to an embodiment of the present disclosure for achieving the aforementioned purpose, a computer program stored in a computer-readable storage medium is disclosed. When one or more processors run the computer program, an action of predicting health status using an electrocardiogram is performed. At this time, the action may include: an action of acquiring electrocardiogram data and cardiac arrest data including detecting whether cardiac arrest occurs in an object of the electrocardiogram data and the time of cardiac arrest; an action of analyzing missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data; and an action of marking the extracted valid data based on whether cardiac arrest occurs and the time of cardiac arrest to generate learning data.

[0019] According to an embodiment of the present disclosure for achieving the aforementioned purpose, a computing device for predicting health status using an electrocardiogram is disclosed. The device includes: a processor including at least one core; a memory including a program code that can be run on the processor; and a network unit, which acquires electrocardiogram data and cardiac arrest data including detecting whether the object of the electrocardiogram data has cardiac arrest and the time of cardiac arrest. At this time, the processor analyzes the missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data, and performs labeling on the extracted valid data based on whether cardiac arrest has occurred and the time of cardiac arrest to generate learning data.

[0020] Effects of the Invention

[0021] The present disclosure may provide a method for learning an artificial intelligence model based on the result of filtering electrocardiogram data reflecting cardiac arrest status, and using the learned model to predict health conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 A block diagram of a computing device according to an embodiment of the present disclosure.

[0023] Figure 2 The block diagram is a diagram showing a process of extracting valid data from electrocardiogram data according to an embodiment of the present disclosure.

[0024] Figure 3 A block diagram illustrating a process of learning a deep learning model based on valid data according to an embodiment of the present disclosure.

[0025] Figure 4 The flowchart is a method for predicting health status by using an electrocardiogram according to an embodiment of the present disclosure.

[0026] Figure 5 The flowchart is a method for predicting health status by using an electrocardiogram according to an optional embodiment of the present disclosure. DETAILED DESCRIPTION

[0027] Hereinafter, embodiments of the present disclosure are described in detail with reference to the accompanying drawings, so that a person of ordinary skill in the technical field to which the present disclosure belongs (hereinafter referred to as a person of ordinary skill) can easily implement the present disclosure. The embodiments proposed in the present disclosure are provided for a person of ordinary skill to utilize or implement the contents of the present disclosure. Therefore, various modifications of the embodiments of the present disclosure are obvious to a person of ordinary skill. That is, the present disclosure can be implemented in a variety of different forms and is not limited to the following embodiments.

[0028] Throughout the specification of the present disclosure, the same or similar reference numerals refer to the same or similar components. In addition, in order to clearly explain the present disclosure, in the drawings, reference numerals of parts that are not related to the explanation of the present disclosure may be omitted.

[0029] The term "or" as used in this disclosure refers to an inclusive "or" and does not refer to an exclusive "or". That is, unless otherwise specified in this disclosure or its meaning is unclear from the context, "X utilizes A or B" should be understood to mean one of the natural inclusive substitutions. For example, unless otherwise specified in this disclosure or its meaning is unclear from the context, "X utilizes A or B" can be interpreted as one of the cases where X utilizes A, or X utilizes B, or X utilizes A and B.

[0030] The term "and / or" used in the present disclosure should be understood to refer to and include all possible combinations of more than one of the listed related concepts.

[0031] The terms "include" and / or "comprising..." used in this disclosure should be understood to refer to the presence of specific features and / or components. However, the terms "include" and / or "comprising..." should be understood to not exclude the presence or addition of one or more other features, other components and / or combinations thereof.

[0032] Unless otherwise specified or the context clearly indicates a singular form in this disclosure, the singular is generally interpreted to include "one or more".

[0033] The term "Nth (N is a natural number)" used in the present disclosure can be understood as being used to distinguish the components of the present disclosure according to a specified standard such as a functional viewpoint, a structural viewpoint, or convenience of description. For example, components that perform different functions in the present disclosure can be distinguished as a first component or a second component. However, components that are substantially the same within the technical concept of the present disclosure but need to be distinguished for the convenience of description can also be distinguished as a first component or a second component.

[0034] The term “acquire” used in the present disclosure refers not only to receiving data through a wired or wireless communication network with an external device or system, but also to generating data in an on-device form.

[0035] In addition, the term "module" or "unit" used in the present disclosure can be understood as a term referring to an independent functional unit for processing computing resources such as a computer-related entity, firmware, software or a part thereof, hardware or a part thereof, a combination of software and hardware, etc. In this case, a "module" or "unit" can be a unit composed of a single element, or a unit represented by a combination or set of multiple elements. For example, in a narrow sense, a "module" or "unit" can refer to a hardware element or a set thereof of a computing device, an application program that executes a specific function of the software, a process (procedure) implemented by running the software, or a set of commands for running a program, etc. In addition, in a broad sense, a "module" or "unit" can refer to the computing device itself that constitutes the system, or an application program running on the computing device, etc. However, the concept is only an example, and the concept of "module" or "unit" can be defined in a variety of ways within the scope that can be understood by ordinary technicians according to the content of the present disclosure.

[0036] 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 collection of software units used to solve a specific problem, or an abstract model of a process for solving a specific problem. For example, a neural network "model" may refer to an entire system implemented by a neural network that has problem-solving capabilities through learning. In this case, the neural network may have problem-solving capabilities by learning to optimize the parameters of connecting nodes or neurons. A neural network "model" may include a single neural network, or a collection of neural networks composed of a plurality of neural networks.

[0037] The above-mentioned terminology is used to help understand the present disclosure. Therefore, it should be noted that, unless the above-mentioned terms are clearly stated as matters for limiting the content of the present disclosure, the content of the present disclosure is not used to limit the meaning of the technical idea.

[0038] Figure 1 A block diagram of a computing device according to an embodiment of the present disclosure.

[0039] The computing device 100 of 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, and may also be a software-based computing environment connected by a communication network. For example, the computing device 100 may be a main server that performs intensive data processing functions and shares resources, and may also be a client that shares resources through interaction with a server. In addition, the computing device 100 may also be a cloud system (cloud system) in which multiple servers and clients interact to comprehensively process data. The above description is only an example related to the type of computing device 100. Therefore, the type of computing device 100 can be configured in various ways within the scope that ordinary technicians can understand based on the content of this disclosure.

[0040] Reference Figure 1 , the computing device 100 of an embodiment of the present disclosure may include a processor 110, a memory 120 and a network unit 130. However, Figure 1 This is only an example, therefore, the computing device 100 may include other configurations for implementing a computing environment. In addition, the computing device 100 may also include only a portion of the disclosed configurations.

[0041] The processor 110 of an embodiment of the present disclosure may be understood as a configuration unit including hardware and / or software for performing computational operations. For example, the processor 110 may perform data processing for machine learning by reading a computer program. The processor 110 may process operations such as processing of input data for machine learning, feature extraction for machine learning, and error calculation based on back propagation. The processor 110 for performing data processing as described above 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 is only an example, and the type of the processor 110 may be configured in a variety of ways within the scope that can be understood by ordinary technicians based on the content of the present disclosure.

[0042] The processor 110 may filter the electrocardiogram data according to the electrocardiogram data and the cardiac arrest data of the object for detecting the electrocardiogram data. The processor 110 may extract valid data that can be used for learning from the electrocardiogram data by analyzing the electrocardiogram data and the cardiac arrest data corresponding thereto. For example, the processor 110 may filter data in the electrocardiogram data that has little learning utility and is difficult to correspond to the prediction result by analyzing whether cardiac arrest occurs and the time of cardiac arrest, etc., included in the cardiac arrest data. The processor 110 may filter data in the electrocardiogram data that has no effect on learning or has a negative impact on learning by analyzing the extent of missing values ​​in the electrocardiogram data. As described above, the processor 110 may filter the electrocardiogram data according to information about cardiac arrest or information missing from the electrocardiogram data itself, so as to screen valid data that can be used for learning.

[0043] The processor 110 may generate learning data for a deep learning model for performing health status prediction based on the filtered electrocardiogram data. The processor 110 performs labeling on the valid data extracted from the electrocardiogram data, and may generate learning data using the labeled valid data. For example, the processor 110 may label the valid data based on information included in the cardiac arrest data corresponding to the valid data. The processor 110 may label information included in the cardiac arrest data, such as whether cardiac arrest occurs and the time of occurrence of cardiac arrest, in the valid data. The processor 110 may determine the valid data labeled in the manner described above as learning data for a model for predicting the occurrence of cardiac arrest.

[0044] The processor 110 can learn a deep learning model for predicting health status based on the generated learning data and using the electrocardiogram. The learning data includes electrocardiogram data and labels as inputs. If the deep learning model outputs a prediction result about the health status according to the input included in the learning data, the processor 110 can adjust the neural network parameters included in the deep learning model by comparing the output and the label. The processor 110 repeatedly performs the adjustment process of the neural network parameters in a manner that minimizes the error between the output and the label to learn the deep learning model. At this time, the deep learning model may include a neural network that extracts features from the electrocardiogram and a neural network that calculates a predicted value about the health status based on the extracted features. For example, the processor 110 inputs valid data of information about cardiac arrest into the deep learning model to be able to output a probability value about the possibility of cardiac arrest (or the possibility of death) in the form of a score between 0 and 100. The processor 110 calculates the error using the score output by the deep learning model and the loss function that takes the label as an input variable, and can adjust the neural network parameters included in the deep learning model in a direction that reduces the error. Processor 110 may learn by repeatedly performing this process in such a way that the deep learning model predicts the occurrence of cardiac arrest.

[0045] The processor 110 can use the learned deep learning model to predict the health status of the object of the detected electrocardiogram data based on the newly acquired electrocardiogram data. The processor 110 can calculate the predicted result of the health status of the object of the detected electrocardiogram data by inputting the electrocardiogram data into the learned deep learning model. At this time, the predicted result of the health status may include a score expressing the degree of deterioration of the overall health status, such as the probability of death of the object of the detected electrocardiogram data. For example, the processor 110 inputs the newly acquired electrocardiogram data into the learned deep learning model to be able to output the probability value of the possibility of cardiac arrest (or the possibility of death) in the form of a score between 0 and 100. At this time, it can be understood that 0 refers to a healthy state with a very low possibility of death, and 100 refers to a dangerous state with the highest possibility of death. That is, the processor 110 can use the deep learning model that has completed learning to calculate a score from the electrocardiogram data that can intuitively understand the overall health status of a person.

[0046] The memory 120 of an embodiment of the present disclosure may be understood as a hardware and / or software configuration unit that stores and manages data processed in the computing device 100. That is, the memory 120 may 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 type of storage medium of a flash memory type, a hard disk type, a multimedia card microtype, 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, and an optical disk. In addition, the memory 120 may also include a database system that controls and manages data under a specified system. The type of the memory 120 is only an example, and therefore, the type of the memory 120 may be configured in various ways within a scope understandable to a person skilled in the art based on the content of the present disclosure.

[0047] The memory 120 can manage by structuring and organizing data required for the processor 110 to perform operations, combinations of data, and program codes (codes) executable in the processor 110. For example, the memory 120 can store medical data received through the network unit 130 described later. The memory 120 can store program codes that allow a neural network model to receive medical data to perform learning, program codes that allow a neural network model to receive medical data to perform inference according to the purpose of use of the computing device 100, and processed data generated by running the program codes.

[0048] The network unit 130 of an embodiment of the present disclosure may be understood as a configuration unit for sending or receiving data through any form of a well-known wired or wireless communication system. For example, the network unit 130 may send or receive data through 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 technology (5G), ultra wide-band wireless communication (ultra wide-band), ZigBee, radio frequency communication, wireless LAN, wireless fidelity, near field communication (NFC) or Bluetooth. The communication system is only an example, and therefore, the wired or wireless communication system used for the network unit 130 to send or receive data may be applicable in a variety of ways in addition to the example.

[0049] The network unit 130 may receive data required for the processor 110 to perform operations through wired or wireless communication with, for example, any system or any client. In addition, the network unit 130 may send data generated by the operation of the processor 110 through wired or wireless communication with, for example, any system or any client. For example, the network unit 130 may receive medical data through communication with a database in a hospital environment, a cloud server that performs operations such as standardization of medical data, a client such as a smart watch, or a medical computing device. The network unit 130 may send output data of the neural network model and intermediate data and processed data derived during the operation of the processor 110 through communication with the aforementioned database, server, client, or computing device.

[0050] Figure 2The block diagram is a diagram showing a process of extracting valid data from electrocardiogram data according to an embodiment of the present disclosure.

[0051] Reference Figure 2 , the computing device 100 of an embodiment of the present disclosure can extract valid data 40 from the electrocardiogram data 10 by analyzing the electrocardiogram data 10 and the cardiac arrest data 20. Specifically, the computing device 100 can determine the cutoff time 39 for filtering the electrocardiogram data by the time of occurrence of cardiac arrest and the clinical evidence of cardiac arrest recorded in the cardiac arrest data 20. In addition, the computing device 100 can extract the valid data 40 from the electrocardiogram data 10 by filtering the electrocardiogram data of the object that has cardiac arrest within the cutoff time 39 in the electrocardiogram data 10. That is, the computing device 100 can calculate the cutoff time 39 based on the information about cardiac arrest to extract the data of the object that has cardiac arrest within a preset time after the cutoff time 39 as the valid data 40. Before reaching the time of death, the electrocardiogram signal changes greatly, and the analysis utility of the prediction close to the time of death is not great. Therefore, the computing device 100 can filter valid data 40 from the electrocardiogram data 10 according to the cut-off time 39 related to the time of occurrence of cardiac arrest to prepare data with analytical value for predicting health conditions.

[0052] For example, the computing device 100 may calculate the candidate time for deriving the cut-off time 39 by analyzing the cardiac arrest occurrence time and the clinical evidence regarding the cardiac arrest recorded in the cardiac arrest data 20. The candidate time may include: a first candidate time 31 calculated based on the clinical judgment of the changes in the electrocardiogram that occurred before the occurrence of cardiac arrest, a second candidate time 33 that is clinically judged as the time when a treatment measure to prevent cardiac arrest can be taken when cardiac arrest is predicted to occur, and a third candidate time 35 corresponding to the error between the cardiac arrest occurrence time recorded in the cardiac arrest data and the actual cardiac arrest occurrence time of the object for which the electrocardiogram data was detected.

[0053] The computing device 100 can calculate the N minutes when the probability of cardiac arrest increases and changes greatly, and the accuracy of the deep learning model analysis is judged to be low as the first candidate time 31. The electrocardiogram detected N minutes before cardiac arrest and N minutes before cardiac arrest usually changes more than the healthy state. In addition, in this part, even if the analysis is performed by the deep learning model, it shows a tendency that the accuracy of the analysis result of the possibility of cardiac arrest is low. Therefore, the computing device 100 can calculate the estimated time when the electrocardiogram change state begins to change to the maximum as the first candidate time 31 based on the clinical judgment of cardiac arrest.

[0054] The computing device 100 may calculate the time M minutes before the predicted time of occurrence of cardiac arrest that the treatment measures to prevent cardiac arrest can be taken as the second candidate time 33. This is because, even if the deep learning model predicts a time shorter than M minutes as the predicted time of occurrence of cardiac arrest, if the time is a short time when the treatment measures to prevent cardiac arrest cannot be taken, the prediction itself performed for the purpose of preventing cardiac arrest in advance is meaningless. Therefore, the computing device 100 may calculate the clinical time when the treatment measures to prevent cardiac arrest can be taken as the second candidate time 33. At this time, the time when the treatment measures to prevent cardiac arrest can be taken can be determined based on the personal information included in the cardiac arrest data and the clinical evidence about cardiac arrest. The computing device 100 may consider the age, gender, etc. of the test person included in the personal information, and analyze that the treatment measures can be taken within a few minutes from the time when cardiac arrest is clinically predicted to occur to prevent cardiac arrest. In addition, the computing device 100 may determine the time derived from the analysis result as the second candidate time 33.

[0055] The computing device 100 may calculate the error L minutes between the time when cardiac arrest actually occurred and the time recorded in the cardiac arrest data by medical personnel or the like as the second candidate time 35. At this time, the time when cardiac arrest actually occurred may be determined based on the electrocardiogram data. The electrocardiogram signal represents cardiac activity, and therefore, the computing device 100 may determine the time when the electrocardiogram signal included in the electrocardiogram data does not change as the time when cardiac arrest actually occurred. Even if the heart of the patient who has cardiac arrest actually stops at 12:24, it may be recorded by medical personnel as having cardiac arrest at 12:30 based on subjective judgment. In this case, even if the electrocardiogram detected after 12:26 is an electrocardiogram taken after the occurrence of cardiac arrest, if it is observed in association with the cardiac arrest data, it is highly likely that it is erroneously analyzed as an electrocardiogram detected 4 minutes before the occurrence of cardiac arrest. Therefore, in order to prevent erroneous analysis based on such an error, the computing device 100 may determine the error between the actual time when cardiac arrest occurred and the time when cardiac arrest occurred included in the cardiac arrest data as the third candidate time 35.

[0056] In order to determine the cut-off time 39 as the time with lower utility analyzed before the occurrence of cardiac arrest, the computing device 100 may compare the sizes of the first candidate time 31, the second candidate time 33, and the third candidate time 35. And, based on the comparison result, the computing device 100 may determine the maximum value among the first candidate time 31, the second candidate time 33, or the third candidate time 35 as the cut-off time 39. For example, the maximum value among N minutes as the first candidate time 31, M minutes as the second candidate time 33, or L minutes as the third candidate time 35 may include the remaining time. Therefore, the computing device 100 may determine the maximum value including the remaining time among the candidate times 31, 33, and 35 as T minutes as the cut-off time 39.

[0057] The computing device 100 may exclude data within the cutoff time 39 from the electrocardiogram data 10, and extract the remaining data as valid data 40. For example, if the cutoff time 39 is determined to be T minutes, the computing device 100 may determine the data within T minutes in the electrocardiogram data 10 as data with low learning utility. The computing device 100 may determine the data after T minutes in the electrocardiogram data 10 as data with high learning utility. In addition, the computing device 100 may determine the data of the subject who has not suffered cardiac arrest in the electrocardiogram data 10 as data with high learning utility. Furthermore, the computing device 100 may determine the data with high learning utility in the electrocardiogram data 10 as valid data 40.

[0058] In addition, although not Figure 2 As shown, the computing device 100 can also extract valid data 40 from the electrocardiogram data 10 by analyzing the missing values ​​of the electrocardiogram data 10. The computing device 100 can analyze the degree of existence of missing values ​​based on the lead itself or the detection time of the electrocardiogram signal included in the electrocardiogram data 10. In addition, the computing device 100 can filter the electrocardiogram data 10 according to the analysis result to generate the valid data 40. For example, in the case where more than one lead itself is missing in the N leads of the electrocardiogram data 10, the computing device 100 can exclude the electrocardiogram data in which more than one lead itself is missing in the N leads from the valid data 40. In the case where there are leads with missing values ​​for more than a specified time in the N leads of the electrocardiogram data 10, the computing device 100 can exclude the electrocardiogram data in which there are leads with missing values ​​for more than a specified time in the N leads from the valid data 40. That is, the computing device 100 can exclude data in which more than one lead itself or part of the N leads is missing from the valid data 40.

[0059] The computing device 100 may use at least one of the methods for generating valid data 40 based on the deadline 39 and the methods for generating valid data 40 based on missing values ​​of the electrocardiogram data 10 according to the user input. For example, in the case where the user who is to generate learning data wants to generate high-purity learning data, the computing device 100 may generate valid data 40 using both the methods for generating valid data 40 based on the deadline 39 and the methods for generating valid data 40 based on missing values ​​of the electrocardiogram data 10 according to the input according to the user's intention.

[0060] Figure 3 A block diagram illustrating a process of learning a deep learning model based on valid data according to an embodiment of the present disclosure.

[0061] Reference Figure 2 and Figure 3 The computing device 100 of an embodiment of the present disclosure may be used for Figure 2The calculated valid data 40 is labeled to generate learning data 50. The computing device 100 can label the valid data 40 based on whether cardiac arrest occurs and the time of cardiac arrest to generate learning data 50. For example, the computing device 100 can label the data of the subjects who have cardiac arrest within 24 hours after the deadline 39 in the valid data 40 as the cardiac arrest group. In addition, the computing device 100 can label the data of the subjects who have not suffered cardiac arrest or have not died within 24 hours after the deadline 39 in the valid data 40 as the normal group. That is, the computing device 100 can also not classify the data whose cardiac arrest time exceeds 24 hours after the deadline 39 into any group to reduce the prediction error of the model in the learning process. In addition, the value of 24 hours is only an example, and the computing device 100 can set the time differently according to the user input.

[0062] The computing device 100 can perform learning of the deep learning model 200 based on the learning data 50 generated by completing the labeling. At this time, the deep learning model 200 can learn in a manner that calculates the possibility of cardiac arrest within 24 hours from the time the electrocardiogram data is received. For example, the computing device 100 can input the electrocardiogram data included in the learning data 50 into the deep learning model 200 to calculate a score representing the possibility of cardiac arrest within 24 hours. At this time, the score value between 0 and 100 represents the possibility of cardiac arrest within 24 hours. The computing device 100 can calculate the error by comparing the score calculated in the deep learning model 200 and the score corresponding to the label through a loss function. The computing device 100 can adjust the neural network parameters of the deep learning model 200 in the direction of reducing the error. The computing device 100 repeatedly performs this process, whereby if the error meets the minimum standard, the learning of the deep learning model 200 can be terminated. In addition, Figure 3 The learning of the deep learning model 200 is described based on guided learning, but according to the neural network structure of the deep learning model 200, the learning of the deep learning model 200 can also be performed based on self-guided learning.

[0063] Figure 4 The flowchart is a method for predicting health status by using an electrocardiogram according to an embodiment of the present disclosure.

[0064] Reference Figure 4, the computing device 100 of an embodiment of the present disclosure may obtain electrocardiogram data and cardiac arrest data including whether the object of the electrocardiogram data has cardiac arrest and the time of cardiac arrest (step S110). In the case where the computing device 100 is a server that performs electrocardiogram interpretation, the computing device 100 may obtain the electrocardiogram data and the cardiac arrest data corresponding thereto by communicating with a device that generates the electrocardiogram data and the cardiac arrest data. In the case where the computing device 100 is a device that can perform electrocardiogram detection and record information related to cardiac arrest, the computing device 100 may generate the electrocardiogram data and the cardiac arrest data according to the signals and information input through the detection unit and the input / output unit.

[0065] The computing device 100 may analyze the missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data (step S120). The computing device 100 may determine a cutoff time for filtering the electrocardiogram data by analyzing the cardiac arrest occurrence time recorded in the cardiac arrest data and the clinical evidence of the cardiac arrest. Furthermore, the computing device 100 may extract valid data by filtering the electrocardiogram data of the subject who has cardiac arrest within the cutoff time from the electrocardiogram data.

[0066] The computing device 100 can perform labeling on the valid data extracted in step S120 based on whether cardiac arrest occurs and the time of cardiac arrest, to generate learning data (step S130). The computing device 100 can classify the data of the subjects who have experienced cardiac arrest and the data of the subjects who have not experienced cardiac arrest or are not dead in the valid data. In addition, the computing device 100 can perform labeling on the data divided into two groups.

[0067] Moreover, although not Figure 4 In the figure, the computing device 100 can learn the deep learning model in a manner that the learning data is input into the deep learning model so that the deep learning model calculates the possibility of cardiac arrest. The description of the learning of the deep learning model will be described in detail. Figure 3 Replace it with the instructions.

[0068] Figure 5 The flowchart is a method for predicting health status by using an electrocardiogram according to an optional embodiment of the present disclosure.

[0069] Reference Figure 5 , the computing device 100 of an embodiment of the present disclosure can obtain electrocardiogram data. In the case where the computing device 100 is a server for interpreting electrocardiograms, the computing device 100 can obtain electrocardiogram data by communicating with a device for detecting electrocardiogram data, such as a wearable device. In the case where the computing device 100 is a device for detecting electrocardiograms, the computing device 100 can generate electrocardiogram data based on signals and information input through a detection unit such as electrodes.

[0070] The computing device 100 can predict the health status of the object detecting the electrocardiogram data by inputting the electrocardiogram data into a pre-learned deep learning model (step S220). The computing device 100 can calculate a score indicating the possibility of cardiac arrest within 24 hours for the object detecting the electrocardiogram data by inputting the deep learning model of the electrocardiogram data. In addition, the computing device 100 can generate a report based on the degree of health condition deterioration analyzed by the score. In the case where the computing device 100 is a server for interpreting electrocardiograms, the computing device 100 can share the report by communicating with a device detecting electrocardiogram data, such as a wearable device. In the case where the computing device 100 is provided with an input and output unit, the computing device 100 can directly output the report through the input and output unit.

[0071] The various embodiments of the present disclosure described above may be combined with additional embodiments and may be changed within the scope that can be understood by ordinary technicians according to the detailed description. It should be understood that the embodiments of the present disclosure are illustrative in all aspects and not restrictive. For example, each component described as a single type can be implemented in a dispersed manner, and the components described as dispersed can also be implemented in a combined form. Therefore, it needs to be explained that all changes or deformations derived from the meaning, scope and equivalent concepts of the invention claimed in the present disclosure are included in the scope of the present disclosure.

Claims

1. A method for predicting health status using an electrocardiogram, performed by a computing device including at least one processor, characterized in that: include: The step of acquiring electrocardiogram data and cardiac arrest data including detecting whether the subject of the electrocardiogram data has cardiac arrest and the time of cardiac arrest occurrence; A step of analyzing missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data; as well as The step of generating learning data by labeling the extracted valid data based on whether cardiac arrest has occurred and the time of cardiac arrest occurrence.

2. The method according to claim 1, characterized in that The step of analyzing the missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data comprises: The step of analyzing the occurrence time of cardiac arrest recorded in the cardiac arrest data and the clinical evidence of cardiac arrest to determine the cut-off time for filtering the electrocardiogram data; and The step of filtering the electrocardiogram data of the subject who has suffered cardiac arrest within the cut-off time from the acquired electrocardiogram data to extract valid data.

3. The method according to claim 2, characterized in that The step of analyzing the cardiac arrest occurrence time recorded in the cardiac arrest data and the clinical evidence of cardiac arrest to determine the cutoff time for filtering the electrocardiogram data comprises: A step of determining as a cutoff time one of a first candidate time calculated based on a clinical judgment of changes in the electrocardiogram that occurred before cardiac arrest occurred, a second candidate time that is clinically judged to be a time when measures to prevent cardiac arrest can be taken when cardiac arrest occurs, or a third candidate time corresponding to an error between the cardiac arrest occurrence time recorded in the cardiac arrest data and the actual cardiac arrest occurrence time of the object for which the electrocardiogram data was detected.

4. The method according to claim 3, characterized in that: The deadline is determined as a maximum value among the first candidate time, the second candidate time, or the third candidate time.

5. The method according to claim 4, characterized in that The step of generating learning data by marking the extracted valid data based on whether cardiac arrest occurs and the time of cardiac arrest comprises: The step of marking the data of the objects that have suffered cardiac arrest within the preset time after the cut-off time in the valid data as a cardiac arrest group, and marking the data of the objects that have not suffered cardiac arrest or have not died within the preset time after the cut-off time in the valid data as a normal group.

6. The method according to claim 1, characterized in that The step of analyzing the missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data comprises: When one or more leads are missing from among the N leads of the electrocardiogram data, the electrocardiogram data in which one or more leads are missing are excluded from the valid data.

7. The method according to claim 1, characterized in that The step of analyzing the missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data comprises: When there is a lead having a missing value for a predetermined time or longer among the N leads of the electrocardiogram data, the step of excluding the electrocardiogram data of the lead having a missing value for a predetermined time or longer among the N leads from the valid data.

8. The method according to claim 1, characterized in that Also includes: The step of inputting the learning data into a deep learning model to learn the deep learning model in such a manner that the deep learning model calculates the possibility of occurrence of cardiac arrest.

9. A method for predicting health status using an electrocardiogram, performed by a computing device including at least one processor, characterized in that: include: The step of acquiring first electrocardiogram data; and The step of inputting the first electrocardiogram data into a pre-learned deep learning model to predict the health status of the object detecting the electrocardiogram data, The deep learning model learns based on learning data generated by labeling valid data extracted from the second electrocardiogram data based on whether cardiac arrest occurs and the time of cardiac arrest occurrence, The valid data is extracted from the second electrocardiogram data according to a result of analyzing missing values ​​of the second electrocardiogram data or cardiac arrest data including detecting whether a subject of the second electrocardiogram data has suffered cardiac arrest and the time of occurrence of cardiac arrest.

10. The method according to claim 9, characterized in that The prediction result derived by inputting the first electrocardiogram into the pre-learned deep learning model includes a score value indicating a health condition of a subject detecting the electrocardiogram data, The closer the score value is to the first threshold, the healthier it is, and the closer the score value is to the second threshold, the less healthy it is.

11. A computer program stored in a computer-readable storage medium, characterized in that: When the computer program is executed by one or more processors, the computer program performs the action of predicting health status by using electrocardiogram. The actions include: The action of acquiring electrocardiogram data and cardiac arrest data including detecting whether a subject of the electrocardiogram data has suffered cardiac arrest and the time of occurrence of cardiac arrest; An action of analyzing missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data; and The operation of generating learning data by labeling the extracted valid data based on whether cardiac arrest has occurred and the time of cardiac arrest occurrence.

12. A computing device for predicting health status using electrocardiogram, characterized in that: include: a processor, comprising at least one core; a memory including program code executable on the processor; and The network unit acquires electrocardiogram data and cardiac arrest data including detecting whether a subject of the electrocardiogram data has cardiac arrest and the time of cardiac arrest, The processor analyzes missing values ​​of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data, and performs labeling on the extracted valid data based on whether cardiac arrest occurs and the time of cardiac arrest occurrence to generate learning data.

Images