Method, program, and apparatus for predicting health status using electrocardiograms.

Abstract

According to an embodiment of the present disclosure, there is disclosed a method, a program, and an apparatus for predicting a health condition using an electrocardiogram, which are performed by a computing device. The method may include the steps of: acquiring electrocardiogram data and cardiac arrest data including whether or not a subject has experienced cardiac arrest and the time of cardiac arrest for which the electrocardiogram data was measured; analyzing missing values of the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data; and generating training data by labeling the extracted valid data based on whether or not a subject has experienced cardiac arrest and the time of cardiac arrest.

Description

Technical Field

[0001] The content of the present disclosure relates to artificial intelligence technology in the medical field. Specifically, it relates to a method of training an artificial intelligence model based on the result of filtering electrocardiogram data reflecting a cardiac arrest state and predicting a health state with the trained model.

Background Art

[0002] An electrocardiogram examination is an examination that records the electrical activity of the heart. The electrocardiogram examination is a relatively simple and cost-effective examination method, which can confirm the health state of the heart and plays an important role in the initial diagnosis and management of heart diseases. For example, the electrocardiogram signal measured through an electrocardiogram examination can be used to confirm whether each part of the heart is operating normally, whether the size and position of the heart are normal, and whether there is damage to the myocardium. And the electrocardiogram signal can be used to diagnose various heart-related problems and predict a person's health state based on such confirmation.

[0003] On the other hand, as artificial intelligence technology develops, attempts to use artificial intelligence technology to predict a health state based on an electrocardiogram have been increasing. The most important part with the introduction of artificial intelligence technology is to ensure appropriate data required for artificial intelligence learning. Since an electrocardiogram records heart activity, cardiac arrest has a significant impact on the interpretation of the electrocardiogram, but most of the prior arts often do not reflect the occurrence and current situation of cardiac arrest in the data secured for artificial intelligence learning.

Summary of the Invention

Problems to be Solved by the Invention

[0004] An object of the present disclosure is to provide a method of training an artificial intelligence model based on the result of filtering electrocardiogram data reflecting a cardiac arrest state and predicting a health state with the trained model.

[0005] However, the issues that this disclosure seeks to address are not limited to those mentioned above, and other issues not mentioned can be clearly understood based on the following description. [Means for solving the problem]

[0006] An embodiment of the present disclosure, which is performed by a computing device, is disclosed as a method for predicting health status using an electrocardiogram. The method includes the steps of: acquiring electrocardiogram data and cardiac arrest data including whether or not cardiac arrest occurred and the duration of cardiac arrest in the subject on whom the electrocardiogram data was measured; extracting valid data from the electrocardiogram data by analyzing missing values ​​in the electrocardiogram data or the cardiac arrest data; and generating training data by performing labeling on the extracted valid data based on whether or not cardiac arrest occurred and the duration of cardiac arrest.

[0007] Alternatively, the step of analyzing missing values ​​in the electrocardiogram data or cardiac arrest data to extract valid data from the electrocardiogram data may include the steps of analyzing the cardiac arrest time and clinical evidence related to the cardiac arrest recorded in the cardiac arrest data to determine a cutoff time for filtering the electrocardiogram data; and filtering the electrocardiogram data of subjects in which cardiac arrest occurred within the cutoff time from the acquired electrocardiogram data to extract valid data.

[0008] Alternatively, the step of determining a cutoff time for filtering the electrocardiogram data by analyzing the cardiac arrest time and clinical evidence related to the cardiac arrest included in the cardiac arrest data may include determining one of the following as the cutoff time: a first candidate time calculated based on a clinical judgment of the changes in the electrocardiogram that appear before cardiac arrest occurs; a second candidate time clinically determined to be the time during which 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 cardiac arrest time recorded in the cardiac arrest data and the actual cardiac arrest time of the subject from whom the electrocardiogram data was measured.

[0009] Alternatively, the cutoff time may be determined to be the largest value among the first candidate time, the second candidate time, or the third candidate time.

[0010] As an alternative, the step of generating training data by labeling the extracted valid data based on whether or not cardiac arrest occurred and the time of cardiac arrest may include the step of labeling data from the valid data in which cardiac arrest occurred within a predetermined time after the cutoff time as the cardiac arrest group, and data from the valid data in which cardiac arrest did not occur or death did not occur within a predetermined time after the cutoff time as the normal group.

[0011] Alternatively, the step of analyzing the missing values ​​or cardiac arrest data of the electrocardiogram data to extract valid data from the electrocardiogram data may include a step of excluding the electrocardiogram data in which one or more leads of the N leads are missing from the valid data.

[0012] Alternatively, the step of analyzing the missing values ​​or cardiac arrest data in the electrocardiogram data to extract valid data from the electrocardiogram data may include a step of excluding from the valid data any electrocardiogram data in which there are leads with missing values ​​for a predetermined period of time or longer in the N leads of the electrocardiogram data.

[0013] Alternatively, the method may further include a step of inputting the training data into a deep learning model and training the deep learning model to calculate the probability of cardiac arrest occurring.

[0014] A method for predicting health status using an electrocardiogram, performed by a computing device according to one embodiment of the present disclosure to address the aforementioned issues, is disclosed. The method may include the steps of: acquiring first electrocardiogram data; and inputting the first electrocardiogram data into a pre-trained deep learning model to predict the health status of the subject on whom the electrocardiogram data was measured. In this case, the deep learning model may be trained on training data generated by performing labeling on valid data extracted from second electrocardiogram data based on the presence or absence of cardiac arrest and the duration of cardiac arrest. The valid data may be extracted from the second electrocardiogram data based on the results of analyzing missing values ​​in the second electrocardiogram data or cardiac arrest data including the presence or absence of cardiac arrest and the duration of cardiac arrest in the subject on whom the second electrocardiogram data was measured.

[0015] Alternatively, the prediction result derived by inputting the first electrocardiogram into the pre-trained deep learning model may include a score value indicating the health status of the subject whose electrocardiogram data was measured. The closer the score value is to a first critical value, the healthier the subject is, and the closer it is to a second critical value, the unhealthier the subject is.

[0016] An embodiment of the present disclosure for achieving the aforementioned problems is disclosed, which includes a computer program stored on a computer-readable storage medium. When the computer program is executed on one or more processors, it performs operations for predicting a health condition using an electrocardiogram. These operations may include: acquiring electrocardiogram data and cardiac arrest data including whether or not cardiac arrest occurred and the duration of cardiac arrest in the subject from whom the electrocardiogram data was measured; analyzing missing values ​​in the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data; and performing labeling on the extracted valid data based on whether or not cardiac arrest occurred and the duration of cardiac arrest to generate training data.

[0017] One embodiment of the present disclosure, which aims to address the aforementioned challenges, discloses a computing device for predicting health status using an electrocardiogram. The device may include a processor having at least one core; memory containing program code executable by the processor; and a network unit for acquiring electrocardiogram data and cardiac arrest data, including whether or not cardiac arrest occurred and the duration of cardiac arrest in the subject from which the electrocardiogram data was measured. In this case, the processor can analyze missing values ​​in the electrocardiogram data or the cardiac arrest data to extract valid data from the electrocardiogram data, and generate training data by labeling the extracted valid data based on whether or not cardiac arrest occurred and the duration of cardiac arrest. [Effects of the Invention]

[0018] This disclosure provides a method for training an artificial intelligence model based on the results of filtering electrocardiogram data to reflect cardiac arrest status, and for predicting health status using the trained model. [Brief explanation of the drawing]

[0019] [Figure 1] This is a block diagram of a computing device according to one embodiment of the present disclosure. [Figure 2] This is a block diagram showing the process of extracting valid data from electrocardiogram data according to one embodiment of the present disclosure. [Figure 3] This is a block diagram showing the process of training a deep learning model based on valid data according to one embodiment of the present disclosure. [Figure 4] This is a flowchart illustrating a method for predicting health status using an electrocardiogram according to one embodiment of the present disclosure. [Figure 5] This flowchart illustrates a method for predicting health status using an electrocardiogram, which is an alternative embodiment of the present disclosure. [Modes for carrying out the invention]

[0020] Hereinafter, embodiments of the present disclosure will be described in detail so that those skilled in the art (hereinafter referred to as those skilled in the art), who have ordinary knowledge in the technical field of the present disclosure, can easily implement them by referring to the attached drawings. The embodiments presented in the present disclosure are provided so that those skilled in the art can utilize or implement the content of the present disclosure. Therefore, various modifications to the embodiments of the present disclosure will be apparent to those skilled in the art. That is, the present disclosure can be embodied in various different forms and is not limited to the following embodiments.

[0021] Throughout the specification of the present disclosure, the same or similar reference numerals refer to the same or similar components. Also, in order to clearly explain the present disclosure, the reference numerals of the parts in the drawings that are not related to the description of the present disclosure may be omitted.

[0022] The term "or" used in the present disclosure is intended to mean an inclusive "or" rather than an exclusive "or". That is, when not specifically specified in the present disclosure or when the meaning is not clear in the context, "X uses A or B" should be understood to mean one of the natural inclusive substitutions. For example, when not specifically specified in the present disclosure or when the meaning is not clear in the context, "X uses A or B" can be interpreted as either X uses A, X uses B, or X uses both A and B.

[0023] The term "and / or" used in the present disclosure should be understood to refer to and include any possible combination of one or more of the listed related concepts.

[0024] [[ID=ID=18]]The terms "comprising" and / or "include" used in the present disclosure should be understood to mean that a specific feature and / or component exists. However, the terms "comprising" and / or "include" should be understood not to exclude the existence or addition of one or more other features, other components, and / or combinations thereof. <ID=

[0025] Where not specifically specified in this disclosure or unless the context makes it clear that it refers to a singular form, the singular should generally be interpreted as including "one or more."

[0026] The term "the Nth (where N is a natural number)" as used in this disclosure can be understood as an expression used to distinguish the components of this disclosure from one another based on predetermined criteria, such as functional, structural, or for ease of explanation. For example, components that perform different functional roles in this disclosure may be distinguished as either the first component or the second component. However, components that are substantially identical within the technical concept of this disclosure but must be distinguished for ease of explanation may also be distinguished as either the first component or the second component.

[0027] As used in this disclosure, the term “acquisition” may be understood to mean not only receiving data through a wired / wireless network with an external device or system, but also generating data in an on-device manner.

[0028] On the other hand, the terms "module" or "unit" as used in this disclosure may be understood to refer to an independent functional unit that processes computing resources, such as a computer-related entity, firmware, software or a part thereof, hardware or a part thereof, or a combination of software and hardware. In this case, a "module" or "unit" may 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" may refer to a hardware element or set thereof of a computing device, an application program that performs a specific function of software, a processing process (procedure) embodied through the execution of software, or a set of instructions for program execution. In a broader sense, a "module" or "unit" may refer to the computing device itself that constitutes a system, or an application executed on the computing device. However, since the above concepts are merely examples, the concept of a "module" or "unit" may be defined in various ways within the scope that a person skilled in the art can understand based on the content of this disclosure.

[0029] As used in this disclosure, the term "model" can be understood as a system embodied using mathematical concepts and language to solve a particular problem, a set of software units for solving a particular problem, or an abstract model relating to a processing step for solving a particular problem. For example, a neural network "model" may refer to any system embodied by a neural network that has problem-solving capabilities through learning. In this case, the neural network can acquire problem-solving capabilities by optimizing the parameters connecting nodes or neurons through learning. A neural network "model" may include a single neural network or a set of neural networks that are combinations of multiple neural networks.

[0030] The explanations of the terms provided above are intended to aid in understanding this disclosure. Therefore, unless explicitly stated to limit the content of this disclosure, the terms are not used to limit the technical ideas contained herein.

[0031] Figure 1 is a block diagram of a computing device according to one embodiment of the present disclosure.

[0032] The computing device 100 according to one embodiment of this disclosure may be a hardware device or part of a hardware device that performs comprehensive data processing and calculations, or it may be a software-based computing environment connected by a communication network. For example, the computing device 100 may be a server that performs intensive data processing functions and shares resources, or it may be a client that shares resources through interaction with a server. Alternatively, the computing device 100 may be a cloud system in which multiple servers and clients interact to process data comprehensively. The above description is merely one example related to the type of computing device 100, and therefore the type of computing device 100 can be configured in a variety of ways within the scope that can be understood by a person skilled in the art based on the content of this disclosure.

[0033] Referring to Figure 1, a computing device 100 according to one embodiment of the present disclosure may include a processor 110, memory 120, and a network unit 130. However, since Figure 1 is merely an example, the computing device 100 may include other configurations to embody a computing environment. Furthermore, only a portion of the disclosed configurations may be included in the computing device 100.

[0034] A processor 110 according to one embodiment of the present disclosure can be understood as a component unit including hardware and / or software for performing computing operations. For example, the processor 110 can read a computer program and perform data processing for machine learning. The processor 110 can handle computational processes such as processing input data for machine learning, feature extraction for machine learning, and error calculation based on backpropagation. Such a processor 110 for performing data processing may include a central processing unit (CPU), a general-purpose graphics processing unit (GPGPU), a tensor processing unit (TPU), an application-specific integrated circuit (ASIC), or a field-programmable gate array (FPGA). The types of processor 110 described above are merely examples, and the types of processor 110 can be configured in a variety of ways that are understandable to those skilled in the art based on the content of the present disclosure.

[0035] The processor 110 can filter electrocardiogram (ECG) data based on the ECG data and the cardiac arrest data of the target for which the ECG data was measured. The processor 110 can analyze the ECG data and the corresponding cardiac arrest data to extract useful data that can be used for learning from the ECG data. For example, the processor 110 can analyze whether or not a cardiac arrest occurred and the duration of the cardiac arrest included in the cardiac arrest data to filter out data from the ECG data that has little learning utility and is difficult to address with prediction results. The processor 110 can analyze the extent to which there are missing values ​​in the ECG data to filter out data from the ECG data that does not affect learning or may even have a negative impact. In this way, the processor 110 can filter the ECG data based on information about cardiac arrest and information about missing values ​​in the ECG data itself to select useful data that can be used for learning.

[0036] The processor 110 can generate training data for a deep learning model that predicts health status based on filtered electrocardiogram data. The processor 110 can label the valid data extracted from the electrocardiogram data and generate training data using the labeled valid data. For example, the processor 110 can label valid data based on information contained in the corresponding cardiac arrest data. The processor 110 can label the valid data with information contained in the cardiac arrest data, such as whether or not a cardiac arrest occurred and the duration of the cardiac arrest. The processor 110 can then decide to use this labeled valid data as training data for a model that predicts the occurrence of cardiac arrest.

[0037] The processor 110 can train a deep learning model that predicts health status using electrocardiograms based on the generated training data. Since the training data includes electrocardiogram data and labels as input, when the deep learning model outputs a prediction result for health status based on the input included in the training data, the processor 110 can compare the output with the labels and adjust the neural network parameters included in the deep learning model. The processor 110 can train the deep learning model by repeatedly performing the process of adjusting the neural network parameters so that the error between the output and the labels is minimized. 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 for health status based on the extracted features. For example, the processor 110 can input valid data labeled with information about cardiac arrest into the deep learning model and output a probability value for the likelihood of cardiac arrest (or the likelihood of death) in the form of a score between 0 and 100. The processor 110 calculates the error using a loss function with the score and labels output through the deep learning model as input variables, and can adjust the neural network parameters included in the deep learning model in a direction that reduces the error. The processor 110 can repeatedly perform this process to train the deep learning model to predict the occurrence of cardiac arrest.

[0038] The processor 110 can use a trained deep learning model to predict the health status of a subject based on newly acquired electrocardiogram (ECG) data. The processor 110 can input the ECG data into the trained deep learning model to calculate a predicted health status of the subject. This predicted health status may include a score that represents the overall degree of health deterioration, such as the likelihood of death for the subject. For example, the processor 110 can input newly acquired ECG data into a trained deep learning model and output a probability value for the likelihood of cardiac arrest (or death) as a score between 0 and 100. A score of 0 can be understood as a very low probability of death, representing a healthy state, while a score of 100 represents the highest probability of death. In other words, the processor 110 can utilize a trained deep learning model to calculate a score from ECG data that provides an intuitive understanding of a person's overall health status.

[0039] A memory 120 according to one embodiment of the present disclosure can be understood as a component unit including hardware and / or software for storing and managing data processed by a 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 type of storage medium from among flash memory type, hard disk type, multimedia card micro type, card type memory, RAM (random access memory), SRAM (static random access memory), ROM (read-only memory), EEPROM (electrically erasable programmable read-only memory), PROM (programmable read-only memory), magnetic memory, magnetic disk, and optical disk. The memory 120 may also include a database system for controlling and managing data in a predetermined manner. The types of memory 120 described above are merely examples; therefore, the types of memory 120 can be configured in a variety of ways that are understandable to those skilled in the art based on the content of this disclosure.

[0040] Memory 120 can structure and organize and manage data, data combinations, and program code (code) that can be executed by the processor 110, which are necessary for the processor 110 to perform calculations. For example, memory 120 can store medical data received through the network unit 130, which will be described later. Memory 120 can store program code that causes the neural network model to perform learning by receiving medical data as input, program code that causes the neural network model to perform inference according to the intended use of the computing device 100 by receiving medical data as input, and processed data generated by the execution of the program code.

[0041] A network unit 130 according to one embodiment of this disclosure can be understood as a component that transmits and receives data through any form of publicly disclosed wired / wireless communication system. For example, the network unit 130 can perform data transmission and reception using wired / wireless communication systems such as local area networks (LANs), wideband code division multiple access (WCDMA), LTE (Long Term Evolution), wireless broadband (WiBro), 5th generation mobile communication (5G), ultra-wide-band wireless communication (UW), ZigBee, radio frequency (RF) communication, wireless LAN, Wi-Fi (Wireless Fidelity), near field communication (NFC), or Bluetooth®. Since the aforementioned communication systems are merely examples, a wide variety of wired / wireless communication systems other than those exemplified above can be applied to the network unit 130 for data transmission and reception.

[0042] The network unit 130 can receive data necessary for the processor 110 to perform calculations via wired / wireless communication with any system or client. The network unit 130 can also transmit data generated through the processor 110's calculations via wired / wireless communication with any system or client. For example, the network unit 130 can receive medical data through communication with databases in a hospital environment, cloud servers performing tasks such as medical data standardization, clients such as smartwatches, or medical computing devices. Through communication with the aforementioned databases, servers, clients, or computing devices, the network unit 130 can transmit output data from neural network models, as well as intermediate and processed data derived during the processor 110's calculation process.

[0043] Figure 2 is a block diagram showing the process of extracting valid data from electrocardiogram data according to one embodiment of the present disclosure.

[0044] Referring to Figure 2, a computing device 100 according to one embodiment of the present disclosure can analyze electrocardiogram data 10 and cardiac arrest data 20 to extract valid data 40 from the electrocardiogram data 10. Specifically, the computer device 100 can analyze the cardiac arrest occurrence time and clinical evidence related to cardiac arrest recorded in the cardiac arrest data 20 to determine a cutoff time 39 for filtering the electrocardiogram data. Then, the computing device 100 can filter the electrocardiogram data of subjects in the electrocardiogram data 10 in which cardiac arrest occurred within the cutoff time 39 and extract it as valid data 40 from the electrocardiogram data 10. In other words, the computing device 100 can calculate the cutoff time 39 based on information regarding cardiac arrest and extract data of subjects in which cardiac arrest occurred within a predetermined time after the cutoff time 39 as valid data 40. Before the time of death, the electrocardiogram signal changes significantly, and predictions close to the time of death do not have great analytical utility. Therefore, the computing device 100 can select valid data 40 from the electrocardiogram data 10 based on a cutoff time 39 related to the time of cardiac arrest, thereby providing data that is valuable for analysis to predict health status.

[0045] For example, the computing device 100 can analyze the cardiac arrest time recorded in the cardiac arrest data 20 and the clinical evidence related to the cardiac arrest to calculate candidate times for determining the cutoff time 39. The candidate times may include a first candidate time 31 calculated based on clinical judgment regarding the changes in the electrocardiogram that appear before cardiac arrest occurs, a second candidate time 33 clinically determined to be the time during which measures to prevent cardiac arrest can be taken when cardiac arrest is predicted, and a third candidate time 35 corresponding to the error between the cardiac arrest time recorded in the cardiac arrest data and the actual cardiac arrest time of the subject from whom the electrocardiogram data was measured.

[0046] The computing device 100 can calculate the first candidate time 31 as the N minutes at which the probability of cardiac arrest increases and the accuracy of analysis through the deep learning model is low due to the large changes. The electrocardiogram measured during the N minutes from N minutes before cardiac arrest occurs until cardiac arrest occurs may generally show a larger range of changes than in a healthy state. Furthermore, even if analysis is performed through the deep learning model for this portion, the accuracy of the analysis results regarding the probability of cardiac arrest may tend to be low. Therefore, the computing device 100 can calculate the first candidate time 31 as the time at which the most significant changes in the electrocardiogram are estimated to begin, based on clinical judgment regarding cardiac arrest.

[0047] The computing device 100 can calculate M minutes as the second candidate time 33, which is the time in which it is determined that measures to prevent cardiac arrest can be taken before the predicted time of cardiac arrest. This is because even if the deep learning model predicts a time shorter than M minutes as the expected time of cardiac arrest, if the time is so short that measures to prevent cardiac arrest are impossible, the prediction itself, which is performed for the purpose of preventing cardiac arrest in advance, becomes meaningless. Therefore, the computing device 100 can calculate the second candidate time 33 as the clinical time in which it is determined that measures to prevent cardiac arrest can be taken. At this time, the time in which it is determined that measures to prevent cardiac arrest can be taken can be determined based on personal information included in the cardiac arrest data and clinical evidence regarding cardiac arrest. The computing device 100 can analyze how many minutes from the time when cardiac arrest is clinically predicted that measures should be taken to prevent cardiac arrest, taking into account the age, gender, etc. of the person being measured, which are included in the personal information. The computing device 100 can then determine the time derived from the analysis result as the second candidate time 33.

[0048] The computing device 100 can calculate L minutes, which is the error between the actual time cardiac arrest occurred and the time recorded in the cardiac arrest data by medical staff, as the second candidate time 35. In this case, the actual time cardiac arrest may be determined based on electrocardiogram data. Since the electrocardiogram signal indicates cardiac activity, the computing device 100 can determine the time when the electrocardiogram signal included in the electrocardiogram data does not change as the actual time cardiac arrest occurred. Even if the heart of a patient who experienced cardiac arrest actually stopped at 12:24, medical staff may subjectively record that the cardiac arrest occurred at 12:30. In such a case, an electrocardiogram measured after 12:26, ​​even though it was taken after the cardiac arrest occurred, is highly likely to be misanalyzed as an electrocardiogram measured 4 minutes before the cardiac arrest occurred when viewed in conjunction with the cardiac arrest data. Therefore, to prevent misanalysis due to such errors, the computing device 100 can determine the error between the actual time cardiac arrest occurred and the time cardiac arrest occurred included in the cardiac arrest data as the third candidate time 35.

[0049] The computing device 100 can compare the magnitudes of the first candidate time 31, the second candidate time 33, and the third candidate time 35 to determine the cutoff time 39, which is the time before cardiac arrest occurs when the analysis is less effective. Based on the comparison results, the computing device 100 can determine the largest value among the first candidate time 31, the second candidate time 33, or the third candidate time 35 as the cutoff time 39. For example, the largest value among the first candidate time 31 (N minutes), the second candidate time 33 (M minutes), or the third candidate time 35 (L minutes) may include the remaining time. Therefore, the computing device 100 can determine the largest value among the candidate times 31, 33, and 35, including the remaining time, as the cutoff time 39 (T minutes).

[0050] The computing device 100 can exclude data within a cutoff time of 39 minutes from the electrocardiogram data 10 and extract the remaining data as valid data 40. For example, if the cutoff time of 39 minutes is determined to be T minutes, the computing device 100 can determine that the data within T minutes in the electrocardiogram data 10 has low learning utility. The computing device 100 can determine that the data after T minutes in the electrocardiogram data 10 has high learning utility. Furthermore, the computing device 100 can determine that the data in the electrocardiogram data 10 in which cardiac arrest did not occur has high learning utility. Finally, the computing device 100 can determine that the data with high learning utility in the electrocardiogram data 10 is valid data 40.

[0051] On the other hand, although not disclosed in Figure 2, the computing device 100 may analyze the missing values ​​in the electrocardiogram data 10 and extract valid data 40 from the electrocardiogram data 10. The computing device 100 can analyze how many missing values ​​exist based on the leads of the electrocardiogram signals included in the electrocardiogram data 10 or the measurement time. Then, the computing device 100 can filter the electrocardiogram data 10 based on the analysis results and generate valid data 40. For example, if one or more leads of the N leads of the electrocardiogram data 10 are missing, the computing device 100 can exclude the electrocardiogram data in which one or more leads of the N leads are missing from the valid data 40. If there are leads in the N leads of the electrocardiogram data 10 that have missing values ​​for a predetermined time or longer, the computing device 100 can exclude the electrocardiogram data in which there are leads with missing values ​​for a predetermined time or longer from the valid data 40. In other words, the computing device 100 can exclude data in which one or more leads of the N leads are missing or partially missing from the valid data 40.

[0052] The computing device 100 can use at least one of the following methods for generating valid data 40 based on the cutoff time 39 or the method for generating valid data 40 based on missing values ​​in the electrocardiogram data 10, based on user input. For example, if a user intending to generate training data wants to generate high-purity training data, the computing device 100 can generate valid data 40 using both the method for generating valid data 40 based on the cutoff time 39 and the method for generating valid data 40 based on missing values ​​in the electrocardiogram data 10, in accordance with the user's intentions.

[0053] Figure 3 is a block diagram showing the process of training a deep learning model based on valid data according to one embodiment of the present disclosure.

[0054] Referring to Figures 2 and 3, a computing device 100 according to one embodiment of the present disclosure can generate training data 50 by performing labeling on the valid data 40 calculated through Figure 2. The computing device 100 can generate training data 50 by performing labeling on the valid data 40 based on whether or not cardiac arrest occurred and the time of cardiac arrest. For example, the computing device 100 can label data in the valid data 40 in which cardiac arrest occurred within 24 hours after the cutoff time 39 as the cardiac arrest group. The computing device 100 can then label data in the valid data 40 in which cardiac arrest did not occur or death did not occur within 24 hours after the cutoff time 39 as the normal group. In other words, the computing device 100 may not classify data in which the time of cardiac arrest exceeds 24 hours after the cutoff time 39 into any group in order to reduce the prediction error of the model during the learning process. On the other hand, the aforementioned 24-hour figure is merely an example, and the computing device 100 can set the relevant time in various ways through user input.

[0055] The computing device 100 can train the deep learning model 200 based on the training data 50 generated after labeling is complete. At this time, the deep learning model 200 may be trained to calculate the probability of cardiac arrest occurring within 24 hours from the time electrocardiogram data is received. For example, the computing device 100 can input the electrocardiogram data included in the training data 50 into the deep learning model 200 and calculate a score indicating the probability of cardiac arrest occurring within 24 hours. At this time, the score may represent the probability of cardiac arrest occurring within 24 hours as a value between 0 and 100. The computing device 100 can calculate the error by comparing the score calculated by the deep learning model 200 with 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 a direction that reduces the error. The computing device 100 can repeatedly perform this process and terminate the training of the deep learning model 200 when the error satisfies a minimum criterion. On the other hand, while Figure 3 explains the training of the deep learning model 200 based on supervised learning, the neural network structure of the deep learning model 200 means that the training of the deep learning model 200 may also be performed based on self-supervised learning.

[0056] Figure 4 is a flowchart illustrating a method for predicting health status using an electrocardiogram according to one embodiment of the present disclosure.

[0057] Referring to Figure 4, a computing device 100 according to one embodiment of the present disclosure can acquire electrocardiogram data and cardiac arrest data including whether or not cardiac arrest occurred in the subject for which the electrocardiogram data was measured and the time of cardiac arrest (S110). If the computing device 100 is a server that performs electrocardiogram interpretation, the computing device 100 can acquire electrocardiogram data and corresponding cardiac arrest data through communication with equipment that generates electrocardiogram data and cardiac arrest data. If the computing device 100 is a device that performs electrocardiogram measurement and recording of cardiac arrest-related information, the computing device 100 can generate electrocardiogram data and cardiac arrest data based on signals and information input through the measurement unit and input / output unit.

[0058] The computing device 100 can analyze missing values ​​or cardiac arrest data in the electrocardiogram data and extract valid data from the electrocardiogram data (S120). The computing device 100 can analyze the cardiac arrest time recorded in the cardiac arrest data and the clinical evidence related to the cardiac arrest to determine a cutoff time for filtering the electrocardiogram data. Then, the computing device 100 can filter the electrocardiogram data of those in which cardiac arrest occurred within the cutoff time and extract valid data.

[0059] The computing device 100 can generate training data by labeling the valid data extracted through step S120 based on whether or not cardiac arrest occurred and the time of cardiac arrest (S130). The computing device 100 can classify the valid data into data for subjects who experienced cardiac arrest and data for subjects who did not experience cardiac arrest or who did not die. The computing device 100 can then perform labeling on the data classified into the two groups.

[0060] Although not disclosed in Figure 4, the computing device 100 can input training data into the deep learning model to train the deep learning model to calculate the probability of cardiac arrest occurring. The explanation regarding the training of the deep learning model will be replaced by the explanation in Figure 3 mentioned above.

[0061] Figure 5 is a flowchart illustrating a method for predicting health status using an electrocardiogram, which is an alternative embodiment of the present disclosure.

[0062] Referring to Figure 5, a computing device 100 according to one embodiment of the present disclosure can acquire electrocardiogram data. If the computing device 100 is a server that performs electrocardiogram interpretation, the computing device 100 can acquire electrocardiogram data through communication with equipment such as a wearable that measures electrocardiogram data. If the computing device 100 is a device that performs electrocardiogram measurement, the computing device 100 can generate electrocardiogram data based on signals and information input through a measurement unit such as electrodes.

[0063] The computing device 100 can input electrocardiogram data into a pre-trained deep learning model and predict the health status of the subject whose electrocardiogram data was measured (S220). Through the deep learning model into which the electrocardiogram data has been input, the computing device 100 can calculate a score that represents the likelihood of cardiac arrest occurring within 24 hours for the subject whose electrocardiogram data was measured. Based on this score, the computing device 100 can analyze how much the health status has deteriorated and generate a report. If the computing device 100 is a server that performs electrocardiogram interpretation, the computing device 100 can share the report with equipment such as a wearable that measures the electrocardiogram data through communication. If the computing device 100 is equipped with an input / output unit, the computing device 100 can output the report directly through the input / output unit.

[0064] The various embodiments of the Disclosure described herein may be combined with additional embodiments and modified in a manner understandable to those skilled in the art in light of the detailed description set forth herein. The embodiments of the Disclosure are to be understood in all respects to be illustrative and not restrictive. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined manner. Accordingly, all modifications or altered forms derived from the meaning, scope, and equivalent concepts of the claims of the Disclosure shall be construed as being within the scope of the Disclosure.

Claims

1. A method for predicting health status using an electrocardiogram, performed by a computing device including at least one processor, A step of acquiring electrocardiogram data and cardiac arrest data including whether or not cardiac arrest occurred in the subject from which the electrocardiogram data was measured, and the duration of the cardiac arrest; The step of analyzing the missing values ​​or cardiac arrest data of the electrocardiogram data to extract valid data from the electrocardiogram data; and The step of generating training data by labeling the extracted valid data based on whether or not cardiac arrest occurred and the duration of cardiac arrest; Includes, The step of analyzing missing values ​​or cardiac arrest data in the electrocardiogram data to extract valid data from the electrocardiogram data is: A step of determining a cutoff time for filtering the electrocardiogram data by analyzing the cardiac arrest time recorded in the cardiac arrest data and the clinical evidence related to the cardiac arrest; and A method comprising the step of filtering the acquired electrocardiogram data for subjects in which cardiac arrest occurred within the cutoff time to extract valid data.

2. The step of determining a cutoff time for filtering the electrocardiogram data by analyzing the cardiac arrest time and clinical evidence related to cardiac arrest included in the cardiac arrest data is: The method according to claim 1, comprising the step of determining a cutoff time to be one of the following: a first candidate time calculated based on a clinical judgment of the changes in the electrocardiogram that appear before cardiac arrest occurs; a second candidate time clinically determined to be the time during which 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 cardiac arrest time recorded in the cardiac arrest data and the actual cardiac arrest time of the subject from which the electrocardiogram data was measured.

3. The aforementioned cutoff time is The method according to claim 2, wherein the value is determined to be the largest value among the first candidate time, the second candidate time, or the third candidate time.

4. The step of generating training data by labeling the extracted valid data based on whether or not cardiac arrest occurred and the duration of cardiac arrest is as follows: The method according to claim 3, further comprising the steps of: labeling data of subjects in the valid data in which cardiac arrest occurred within a predetermined time after the cutoff time as the cardiac arrest group; and labeling data of subjects in the valid data in which cardiac arrest did not occur or death did not occur within a predetermined time after the cutoff time as the normal group.

5. The step of analyzing missing values ​​or cardiac arrest data in the electrocardiogram data to extract valid data from the electrocardiogram data is: The method according to claim 1, further comprising the step of removing from the valid data the electrocardiogram data in which one or more of the N leads are missing if one or more of the N leads are missing.

6. The step of analyzing missing values ​​or cardiac arrest data in the electrocardiogram data to extract valid data from the electrocardiogram data is: The method according to claim 1, further comprising the step of removing from the valid data any N leads in the electrocardiogram data in which there are missing values ​​for a predetermined period of time or longer.

7. The method according to claim 1, further comprising the step of inputting the training data into a deep learning model and training the deep learning model so that it calculates the probability of cardiac arrest occurring.

8. A method for predicting health status using an electrocardiogram, performed by a computing device including at least one processor, The stage of acquiring the first electrocardiogram data; and The step of inputting the first electrocardiogram data into a pre-trained deep learning model and predicting the health status of the subject from whom the first electrocardiogram data was measured; Includes, The aforementioned deep learning model, The system is trained based on training data generated by labeling the valid data extracted from the second electrocardiogram data based on whether or not cardiac arrest occurred and the duration of cardiac arrest. The aforementioned valid data is Based on the results of analyzing the missing values ​​of the second electrocardiogram data or the cardiac arrest data including whether or not cardiac arrest occurred in the subject for whom the second electrocardiogram data was measured, and the duration of cardiac arrest, the following is extracted from the second electrocardiogram data: The analysis of the missing values ​​or cardiac arrest data is as follows: A step of determining a cutoff time for filtering the second electrocardiogram data by analyzing the cardiac arrest time recorded in the cardiac arrest data and the clinical evidence regarding the cardiac arrest; and A method comprising the step of filtering the acquired electrocardiogram data for subjects in which cardiac arrest occurred within the cutoff time to extract valid data.

9. The prediction result derived by inputting the first electrocardiogram data into the pre-trained deep learning model is: The first electrocardiogram data includes a score value indicating the health status of the subject from which the data was measured, The method according to claim 8, wherein the score value is closer to a first critical value, indicating a healthier state, and is closer to a second critical value, indicating an unhealthier state.

10. A computer program stored on a computer-readable storage medium, wherein, when the computer program is executed on one or more processors, it performs an operation to predict a health condition using an electrocardiogram. The aforementioned operation is, An operation to acquire electrocardiogram data, and cardiac arrest data including whether or not cardiac arrest occurred in the subject from which the electrocardiogram data was measured, and the duration of the cardiac arrest; An operation to extract valid data from the electrocardiogram data by analyzing missing values ​​or cardiac arrest data from the electrocardiogram data; and The operation involves labeling the extracted valid data based on whether or not cardiac arrest occurred and the duration of cardiac arrest to generate training data; Includes, The operation of analyzing missing values ​​or cardiac arrest data in the electrocardiogram data to extract valid data from the electrocardiogram data is: An operation to determine a cutoff time for filtering the electrocardiogram data by analyzing the cardiac arrest time recorded in the cardiac arrest data and the clinical evidence related to the cardiac arrest; and A computer program that includes the operation of filtering the acquired electrocardiogram data for subjects in which cardiac arrest occurred within the cutoff time to extract valid data.

11. A computing device for predicting health status using an electrocardiogram, A processor containing at least one core; Memory containing program code executable by the processor; and A network unit for acquiring electrocardiogram data and cardiac arrest data, including whether or not cardiac arrest occurred in the subject from which the electrocardiogram data was measured, and the duration of the cardiac arrest; Includes, The aforementioned processor, The missing values ​​or cardiac arrest data of the electrocardiogram data are analyzed to extract valid data from the electrocardiogram data. The extracted valid data is labeled based on the presence or absence of cardiac arrest and the duration of cardiac arrest to generate training data. The process of analyzing missing values ​​or cardiac arrest data in the electrocardiogram data to extract valid data from the electrocardiogram data is as follows: A process for determining a cutoff time for filtering the electrocardiogram data by analyzing the cardiac arrest time recorded in the cardiac arrest data and the clinical evidence related to the cardiac arrest; and An apparatus comprising the process of filtering the acquired electrocardiogram data for subjects in which cardiac arrest occurred within the cutoff time to extract valid data.

Images