Method, computer program, and device for continuously measuring body condition on basis of deep learning

Abstract

To provide a method of individually predicting biological information on body characteristics and pathological information on diseases based on electrocardiogram data, and combining these pieces of information to measure a body condition in a continuous numerical way.SOLUTION: According to one embodiment of the present disclosure, disclosed are a method, a computer program, and a device for continuously measuring a body condition on the basis of deep learning, performed by a computing device. The method comprises the steps of: acquiring electrocardiogram data; and using a pretrained neural network model to infer, on the basis of the electrocardiogram data, a physical condition corresponding to the onset of a disease or the progress of the disease of a subject whose electrocardiogram data have been measured, wherein the neural network model may be trained on the basis of at least one of a first feature associated with biological information representing body characteristics having a correlation with the disease, and a second feature associated with pathological information reflecting the progression of the disease.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to deep learning technology in the medical field, and more particularly to a method, computer program, and device for measuring continuous physical conditions based on deep learning to indicate physical characteristics for disease as continuous numerical values.

[0002] [Background technology]

[0003] An electrocardiogram (ECG) signal is a graphical representation of the electrical activity of the heart during each beating cycle. Therefore, we can observe the structural and functional aspects of the heart through the ECG signal. Therefore, ECG signals are used to diagnose various heart diseases, including arrhythmias, myocardial infarction, and other heart disorders.

[0004] Common physical illnesses, including heart disease, are characterized by a continuously changing physical condition that deteriorates to the point where it reaches a standard value established for diagnosing the disease. In other words, the physical condition changes continuously like an analog rather than a digital value of 0 or 1, and the presence or absence of disease is determined by an artificially established standard value. For example, in the case of heart disease, the condition of the coronary arteries that supply blood to the heart changes continuously, and depending on the degree of narrowing of the coronary arteries, diseases such as arteriosclerosis, angina pectoris, and myocardial infarction are determined.

[0005] Various deep learning-based technologies have been developed to predict physical illnesses. However, most of these conventional technologies are based on artificially set criteria to determine the presence or absence of a disease. That is, the presence or absence of a disease is determined by classifying a continuously changing physical condition according to arbitrary criteria. Deep learning models trained using these criteria can only predict the presence or absence of a disease, but cannot predict the overall physical condition related to the disease.

[0006] The purpose of disease prediction is to predict the probability of disease onset and prevent it. Therefore, in the field of disease prediction, it is important not only to predict the presence or absence of a disease, but also to accurately predict the physical state associated with the disease. In other words, we need to predict the continuously changing physical state associated with the disease.

[0007] Summary of the Invention [Problem to be solved by the invention]

[0008] The present disclosure has been devised in response to the above-mentioned background art, and aims to provide a method for individually predicting biological information regarding physical characteristics and pathological information regarding diseases based on electrocardiogram data, and combining the information to measure physical conditions as continuous numerical values.

[0009] However, the problems to be solved by the present disclosure are not limited to those mentioned above, and other problems not mentioned will be clearly understood from the description below.

[0010] [Means for solving the problem]

[0011] To achieve the above object, one embodiment of the present disclosure provides a method for measuring continuous physical conditions based on deep learning, which is executed by a computing device, and includes the steps of acquiring electrocardiogram data and inferring, based on the electrocardiogram data, a physical condition corresponding to the onset of a disease or the progression of the disease in a subject whose electrocardiogram data is measured, using a pre-trained neural network model, wherein the neural network model is trained based on at least one of a first feature for biological information indicating the physical characteristic correlated with the disease or a second feature for pathological information reflecting the progression of the disease.

[0012] Alternatively, the neural network model may include a first sub-model trained to output the first feature based on the electrocardiogram data, and the first sub-model may be configured according to the number of factors included in the biological information to individually output a numerical value for at least one factor.

[0013] Alternatively, the neural network model may further include a second sub-model trained to output the second feature based on the electrocardiogram data, and the second sub-model may be configured according to the number of factors included in the pathological information so as to individually output a numerical value for at least one factor.

[0014] Alternatively, the neural network model may further include a third sub-model trained to numerically represent a physical condition that continuously changes depending on the onset or progression of the disease based on a first feature that is the output of the first sub-model and a second feature that is the output of the second sub-model.

[0015] Alternatively, the third sub-model may receive a third feature generated by combining the first feature and the second feature based on a weight determined by the type of disease and output the numerical value.

[0016] Alternatively, each of the first sub-model and the second sub-model may be trained based on self-supervised learning performed using training data including samples for which no labels have been assigned.

[0017] Alternatively, the disease may include a vascular disease.

[0018] Alternatively, the biological information may include at least one of age, sex, height, or weight as a physical characteristic factor related to coronary artery disease included in the cardiovascular disease.

[0019] Alternatively, the pathological information may include at least one of the following pathological characteristic factors reflecting the degree of progression of coronary artery disease included in the cardiovascular disease: presence or absence of myocardial infarction, degree of vascular calcification, risk of thrombus, intravascular velocity of the coronary artery, or degree of stenosis of the coronary artery.

[0020] According to one embodiment of the present disclosure, there is provided a computer program stored on a computer-readable storage medium. When executed by one or more processors, the computer program performs an operation of continuously measuring a physical condition based on deep learning. The operation includes: acquiring electrocardiogram data; and inferring a physical condition corresponding to the onset of a disease or the progression of the disease in a subject whose electrocardiogram data is measured, based on the electrocardiogram data, using a pre-trained neural network model, wherein the neural network model is trained based on at least one of a first feature of biological information indicating the physical characteristic correlated with the disease or a second feature of pathological information reflecting the progression of the disease.

[0021] To achieve the above object, one embodiment of the present disclosure provides a computing device for measuring continuous physical conditions based on deep learning, the device including: a processor including at least one core; a memory including program code executable by the processor; and a network unit for acquiring electrocardiogram data, wherein the processor infers a physical condition corresponding to the onset or progression of a disease in the subject from which the electrocardiogram data is measured, based on the electrocardiogram data, using a neural network model trained based on at least one of a first feature of biological information indicating a physical characteristic of the subject correlated with a disease or a second feature of pathological information reflecting a degree of progression of the disease.

[0022] [Effects of the Invention]

[0023] The present disclosure can provide a method for individually inferring physical factors and pathological factors for a disease using electrocardiogram data, and for preparing for earlier or later stages of a disease by interpretably grasping factors that affect the physical state. The present disclosure can also provide a method for preparing for earlier or later stages of a disease by indicating a patient's physical state related to the disease to be predicted as a continuous numerical value.

[0024] [Brief explanation of the drawings]

[0025] [Figure 1] FIG. 1 is a block diagram of a computing device according to one embodiment of the present disclosure.

[0026] [Figure 2] FIG. 1 is a block diagram illustrating a neural network model according to one embodiment of the present disclosure.

[0027] [Figure 3] FIG. 2 is a block diagram showing the internal configuration of a neural network model according to an embodiment of the present disclosure.

[0028] [Figure 4] 1 is a flowchart illustrating an inference method for a neural network model according to an embodiment of the present disclosure.

[0029] [Figure 5] 1 is a flowchart illustrating an inference method for sub-models constituting a neural network model according to an embodiment of the present disclosure.

[0030] DETAILED DESCRIPTION OF THE INVENTION

[0031] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the accompanying drawings so that those skilled in the art can easily implement the present disclosure. The embodiments presented in this disclosure are provided to enable those skilled in the art to use or practice the contents 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 may be embodied in various different forms and is not limited to the following embodiments.

[0032] 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, reference numerals of parts that are not relevant to the explanation of the present disclosure may be omitted from the drawings.

[0033] The term "or" as used in this disclosure is intended to mean an inclusive "or" rather than an exclusive "or." That is, unless otherwise specified or otherwise clear from the context in this disclosure, "X uses A or B" should be understood to mean one of the natural inclusive permutations. For example, unless otherwise specified or otherwise clear from the context in this disclosure, "X uses A or B" can be interpreted as either X uses A, X uses B, or X uses both A and B.

[0034] The term "and / or" as used in this disclosure must be understood to indicate and include all possible combinations of one or more of the associated listed concepts.

[0035] The terms "comprises" and / or "comprising" as used in this disclosure should be understood to mean that the specified features and / or components are present. However, the terms "comprises" and / or "comprising" should not be understood to exclude the presence or addition of one or more other features, other components and / or combinations thereof.

[0036] In this disclosure, unless otherwise specified or clear from the context as referring to the singular form, the singular should generally be construed as including "one or more."

[0037] The term "Nth (N is a natural number)" used in this disclosure can be understood as an expression used to distinguish components of the present disclosure from one another based on a predetermined criterion, such as functional, structural, or convenience of description. For example, in this disclosure, components that perform different functional roles can be classified as a first component or a second component. However, components that are substantially identical within the technical concept of the present disclosure but must be distinguished for convenience of description can also be classified as a first component or a second component.

[0038] The term "acquire" as used in this disclosure may be understood to mean not only receiving data from an external device or system via a wired or wireless communication network, but also generating data in an on-device form.

[0039] Meanwhile, the terms "module" or "unit" 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 portion thereof, hardware or a portion thereof, or a combination of software and hardware. Here, a "module" or "unit" may refer to a unit composed of a single element or a unit expressed as a combination or collection of multiple elements. For example, as a concept of connotation, a "module" or "unit" may refer to a hardware element or a collection of hardware elements of a computing device, an application program that performs a specific software function, a processing procedure implemented by executing software, or a collection of instructions for executing a program. Furthermore, as a broad concept, a "module" or "unit" may refer to a computing device itself that constitutes a system, or an application executed on a computing device. However, the above concepts are merely examples, and the concepts of a "module" or "unit" may be defined in various ways within the scope of what one skilled in the art can understand based on the contents of this disclosure.

[0040] The term "model" as used in this disclosure may be understood as a system implemented using mathematical concepts and language to solve a specific problem, a set of software units for solving a specific problem, or an abstract model of a processing process for solving a specific problem. For example, a neural network "model" may refer to a system implemented as a neural network that has problem-solving capabilities through learning. Here, a neural network may have problem-solving capabilities by optimizing parameters connecting nodes or neurons through learning. A neural network "model" may include a single neural network or a neural network ensemble that combines multiple neural networks.

[0041] The explanations of the above terms are intended to aid in understanding the present disclosure. Therefore, unless the above terms are explicitly stated as matters limiting the contents of the present disclosure, care should be taken not to use them to limit the technical ideas of the contents of the present disclosure.

[0042] FIG. 1 is a block diagram of a computing device according to one embodiment of the present disclosure.

[0043] The computing device 100 according to an embodiment of the present disclosure may be a hardware device or part of a hardware device that performs comprehensive data processing and calculations, or may be a software-based computing environment connected via a communication network. For example, the computing device 100 may be a server that performs intensive data processing functions and shares resources, or a client that shares resources by interacting with the server. The computing device 100 may also be a cloud system in which multiple servers and clients interact with each other to comprehensively process data. The above description is merely an example of a type of computing device 100, and various types of computing devices 100 may be configured within the scope of what one skilled in the art would understand based on the contents of this disclosure.

[0044] 1, a computing device 100 according to an embodiment of the present disclosure may include a processor 110, a memory 120, and a network unit 130. However, since FIG. 1 is merely an example, the computing device 100 may include other components for implementing a computer environment. Also, the computing device 100 may include only some of the disclosed components.

[0045] The processor 110 according to an embodiment of the present disclosure may be understood as a component including hardware and / or software for performing computing operations. For example, the processor 110 may read a computer program to perform data processing for machine learning. The processor 110 may process operations such as input data processing for machine learning, feature extraction for machine learning, and error calculation based on backpropagation. The processor 110 for performing such data processing may include a central processing unit (CPU), a general-purpose graphics processing unit (GPGPU), a tensor processing unit (TPU), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), etc. The types of processor 110 described above are merely examples, and various types of processor 110 may be configured within the scope of what one skilled in the art would understand based on the present disclosure.

[0046] The processor 110 can use a neural network model based on the electrocardiogram data to measure a continuous physical state of the subject whose electrocardiogram data was measured. The processor 110 can receive the electrocardiogram data and train the neural network model to estimate a physical state associated with a disease, such as the presence or absence of a disease or a physical state corresponding to the progression of a disease. Here, the presence or absence of a disease may be a discrete value, and the progression of a disease may be a continuous value that changes over time. The processor 110 can use the neural network model to infer a continuous physical state that indicates or affects a disease.

[0047] According to the present disclosure, the processor 110 not only predicts whether a disease will occur as a discrete value, but also predicts the physical condition of the disease as it progresses continuously, thereby accurately predicting the likelihood of the disease occurring. Therefore, the processor 110 of the present disclosure allows us to establish detailed disease prevention and treatment plans.

[0048] The processor 110 can individually train multiple sub-models to predict a physical condition. For example, the processor 110 can use electrocardiogram data to train a first sub-model that outputs a first feature for biological information indicating a physical characteristic of a subject measured in the electrocardiogram data. The processor 110 can then use the electrocardiogram data to train a second sub-model that outputs a second feature for pathological information that affects physical changes caused by a disease, i.e., the progression of the disease. The processor 110 can then train a third sub-model that infers a physical condition associated with the disease, using the first feature output from the first sub-model and the second feature output from the second sub-model.

[0049] The first feature may include at least one factor included in the biological information and a numerical value corresponding to the factor. The biological factor may vary depending on the disease. For example, the biological factor may include age, height, weight, etc. The second feature may include at least one factor included in the pathological information and a numerical value corresponding to the factor. The pathological factor may vary depending on the disease. For example, in the case of cardiovascular disease, the pathological factor may include the presence or absence of myocardial infarction, the degree of vascular calcification, the risk of thrombosis, the intravascular velocity of the coronary artery, the degree of coronary artery stenosis, etc.

[0050] The processor 110 can input learning data including samples and labels corresponding to the samples to the first sub-model and the second sub-model to perform learning based on supervised learning. In this case, the samples input to the first sub-model can be electrocardiogram data, and the labels can be first features. The samples input to the second sub-model can be electrocardiogram data, and the labels can be second features. The processor 110 can also input learning data including samples for which no labels are assigned to the first sub-model and the second sub-model to perform learning based on self-supervised learning. In this case, the first sub-model can extract a first feature from the electrocardiogram data, and the second sub-model can extract a second feature from the electrocardiogram data. The factors extracted by the first sub-model and the second sub-model can vary depending on the type of disease.

[0051] Each of the neural network model, the first sub-model, the second sub-model, and the third sub-model may include at least one neural network, which may include, but is not limited to, at least one of neural networks such as a deep neural network (DNN), a recurrent neural network (RNN), a bidirectional recurrent deep neural network (BRDNN), a multilayer perceptron (MLP), a convolutional neural network (CNN), a transformer, etc.

[0052] According to the present disclosure, the processor 110 can individually learn and infer neural network models corresponding to factors that may act as risk factors for a disease. Thus, we can accurately grasp the degree to which each factor affects a disease through the processor 110. The processor 110 can then accurately target factors that affect a disease, thereby providing basic information for the prevention and treatment of the disease.

[0053] The memory 120 according to an embodiment of the present disclosure may be understood as a component including hardware and / or software for storing and managing data processed by the computing device 100. That is, the memory 120 may store any type of data generated or determined by the processor 110 and any type of data received by the network unit 130. For example, the memory 120 may include at least one type of storage medium selected from the group consisting of flash memory, hard disk, multimedia card micro, card-type memory, random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic memory, magnetic disk, and optical disk. The memory 120 may also include a database system that manages data in a predetermined manner. The types of memory 120 described above are merely examples, and various configurations of the memory 120 are possible within the scope of what would be understood by one skilled in the art based on the present disclosure.

[0054] The memory 120 may structure and organize and manage data, data combinations, and program code executable by the processor 110 required for the processor 110 to perform calculations. For example, the memory 120 may store medical data received via the network unit 130 (described below). The memory 120 may store program code for operating a neural network model to receive medical data and perform learning, program code for operating the neural network model to receive medical data and perform inference according to the intended use of the computing device 100, and processed data generated by executing the program code.

[0055] The network unit 130 according to an embodiment of the present disclosure may be understood as a component that transmits and receives data via any type of known wired or wireless communication system. For example, the network unit 130 may transmit and receive data using a wired or wireless communication system such as a local area network (LAN), wideband code division multiple access (WCDMA), long term evolution (LTE), wireless broadband internet (WiBro), 5G, ultra wide-band, ZigBee, radio frequency (RF) communication, wireless LAN, wireless fidelity (WiFi), near field communication (NFC), or Bluetooth. The above-described communication systems are merely examples, and various wired and wireless communication systems for transmitting and receiving data by the network unit 130 may be applied in addition to the above examples.

[0056] The network unit 130 may receive data necessary for the processor 110 to perform calculations via wired or wireless communication with any system or any client. The network unit 130 may also transmit data generated by calculations by the processor 110 via wired or wireless communication with any system or any client. For example, the network unit 130 may receive medical data via communication with a database in a hospital environment, a cloud server that performs tasks such as standardizing medical data, or a computing device. The network unit 130 may transmit output data of the neural network model, and intermediate data and processed data derived during the calculation process of the processor 110 via communication with the database, server, computing device, or the like.

[0057] FIG. 2 is a block diagram illustrating a neural network model according to one embodiment of the present disclosure.

[0058] Referring to FIG. 2, a neural network model 200 receives electrocardiogram data 300 and can output a physical condition 400 related to the disease of the subject whose electrocardiogram data 300 was measured as a continuous numerical value.

[0059] In this specification, electrocardiogram (ECG) data may include an electrocardiogram signal that measures electrical signals generated in the heart and determines whether or not there is an abnormality in the conduction system from the heart to the electrodes, thereby determining whether or not there is a disease. The electrocardiogram data 300 may be acquired from an electrocardiogram measuring device or via a network.

[0060] The neural network model 200 may include a first sub-model 210 that outputs a first feature for biological information related to a disease, a second sub-model 220 that outputs a second feature for pathological information related to the disease, and a third sub-model 230 that uses the output values ​​of the first sub-model 210 and the second sub-model 220 to output a continuously changing physical condition 400 as a numerical value for diagnosing the disease.

[0061] Each of the first sub-model 210 and the second sub-model 220 may be configured with a plurality of sub-models depending on the information to be output. For example, the first sub-model 210 may output a first feature corresponding to each physical characteristic factor indicating a biological feature related to a disease from the electrocardiogram data 300. Here, the number of first sub-models 210 may correspond to the number of physical characteristic factors. Each of the plurality of first sub-models 210 may individually output a numerical value for the physical characteristic factor. For example, the second sub-model 220 may output a second feature corresponding to each pathological characteristic factor determined according to the degree of progression of a disease from the electrocardiogram data 300. Here, the number of second sub-models 220 may correspond to the number of pathological characteristic factors. Each of the plurality of second sub-models 220 may individually output a numerical value for the pathological characteristic factor.

[0062] Meanwhile, the number of risk factors associated with a disease may vary depending on the type of disease. Therefore, the neural network model according to the present disclosure can configure the first and second sub-models according to the number of factors, allowing us to manage the first and second sub-models in a modular manner.

[0063] The third sub-model 230 may output a physical condition 400 related to a disease as a numerical value using the first feature output from the first sub-model 210 and the second feature output from the second sub-model 220. The third sub-model 230 may output the physical condition 400 based on a third feature generated by combining the first and second features. Here, the first and second features may be weighted according to the type of disease and combined as the third feature. In other words, the processor 110 may adjust the weight for the combination of the first and second features according to the type of disease. The third sub-model 230 may predict the physical condition 400, which continuously changes depending on the type of disease, based on the first and second features whose weights are adjusted according to the type of disease. The value output from the third sub-model 230 may be a continuous numerical value indicating the physical condition 400 for the disease. Therefore, medical staff may determine whether or not a disease has occurred or understand the progression of the disease based on the numerical value.

[0064] That is, the neural network model 200 according to the present disclosure can predict a physical condition by comprehensively grasping disease-related factors, rather than simply predicting whether or not a disease will occur based on an artificially set standard. Using such a neural network model 200, we can accurately grasp complex changes in the physical condition that occur as a result of various disease-related factors affecting the body.

[0065] FIG. 3 is a block diagram illustrating the internal configuration of a neural network model according to an embodiment of the present disclosure.

[0066] Referring to FIG. 3, the neural network model 200 may include a plurality of first sub-models 210 trained to output a first feature 211, a plurality of second sub-models 220 trained to output a second feature 221, and a third sub-model 230 trained to combine the first feature 211 and the second feature 221 to output a physical state 400.

[0067] The first feature 211 may include factors included in the biological information and numerical values ​​corresponding to the factors, and the second feature 221 may include factors included in the pathological information and numerical values ​​corresponding to the factors.

[0068] The type of physical characteristic factor that the first sub-model 210 individually learns may be determined by a label included in the learning data, or may be directly extracted by the first sub-model 210. Similarly, the type of pathological characteristic factor that the second sub-model 220 individually learns may be determined by a label included in the learning data, or may be directly extracted by the second sub-model 220. That is, depending on the learning method, the first sub-model 210 and the second sub-model 220 may output only a numerical value corresponding to the factor, or may output a numerical value and the corresponding factor together.

[0069] In the following, an example will be described in which the neural network model 200 operates to measure the physical condition 400 for cardiovascular disease, particularly coronary artery disease, but the type of disease to which the present disclosure is applicable is not limited to this. Risk factors related to the induction of coronary artery disease are widely known. Even risk factors that are not causally related to coronary artery disease can be used as an important tool for clarifying the cause of the disease or preventing the disease because they are correlated with the disease.

[0070] For example, physical characteristics that affect coronary artery disease may include age, sex, height, weight, etc., and may further include family or personal history of coronary artery disease. Pathological characteristics that affect coronary artery disease may include the presence or absence of myocardial infarction, the degree of vascular calcification, thrombus risk, coronary artery intravascular velocity, the degree of coronary artery stenosis, etc., and may further include blood glucose, blood pressure, cholesterol level, triglyceride level, whether or not the patient is obese, or whether or not the patient is a smoker, etc.

[0071] The neural network model 200 according to the present disclosure can individually learn and infer the influence of risk factors related to the induction of coronary artery disease on coronary artery disease using relatively easily obtainable electrocardiogram data 300. Furthermore, the neural network model 200 according to the present disclosure can learn by self-supervised learning to extract factors related to the induction of coronary artery disease on its own and infer the influence of each factor on coronary artery disease.

[0072] When the first sub-model 210 and the second sub-model 220 are trained by supervised learning, the multiple first sub-models 210 receive electrocardiogram data 300 and are trained to infer age, sex, height, and weight from the electrocardiogram data 300, respectively.

[0073] The plurality of first sub-models 210 may output a numerical value for age, a numerical value for gender, a numerical value for height, and a numerical value for weight, respectively. These numerical values ​​may indicate the impact on coronary artery disease. The plurality of second sub-models 220 receive electrocardiogram data 300 and are trained to infer the presence or absence of myocardial infarction, the degree of vascular calcification, the safety of thrombus, the intravascular velocity of coronary arteries, and the degree of stenosis of coronary arteries, respectively, from the electrocardiogram data 300. The plurality of second sub-models 220 may output a numerical value for myocardial infarction, a numerical value for vascular calcification, the safety of thrombus, the intravascular velocity of coronary arteries, and the stenosis of coronary arteries, respectively. These numerical values ​​may indicate the impact on coronary artery disease.

[0074] The third sub-model 230 may receive a third feature generated based on the values ​​output from the first sub-model 210 and the second sub-model 220. The third feature is a value combined by the processor 110, and the processor 110 may combine the values ​​output from the first sub-model 210 and the second sub-model 220 by assigning weights to them depending on the type of disease. For example, if it is determined that there is a high correlation between a specific disease and physical characteristic factors, the processor 110 may assign a high weight to the value output from the first sub-model 210. Conversely, if it is determined that the physical characteristic factors have a low influence on a specific disease and that the incidence rate due to pathological characteristic factors is high, the processor 110 may assign a high weight to the value output from the second sub-model 220.

[0075] The third sub-model 230 can output a physical condition 400 for coronary artery disease as a numerical value using the third feature. The output numerical value can be interpreted as a numerical value indicating the health of the coronary arteries, the probability of coronary artery disease, the degree of angina pectoris, or the possibility of myocardial infarction. Medical staff can interpret the numerical value and diagnose angina pectoris for the subject whose electrocardiogram data 300 was measured, or create a treatment plan such as administering medication or performing surgery. Alternatively, the medical staff can establish and recommend a disease prevention plan.

[0076] FIG. 4 is a flowchart illustrating an inference method for a neural network model according to one embodiment of the present disclosure.

[0077] 4, a computing device 100 according to an embodiment of the present disclosure may acquire electrocardiogram data (S110). The computing device 100 may acquire the electrocardiogram data from an electrocardiogram measuring device or via a network.

[0078] The computing device 100 can use the pre-trained neural network model to infer a physical condition corresponding to the onset or progression of a disease of the subject whose electrocardiogram data was measured based on the electrocardiogram data (S120). The computing device 100 can pre-train the neural network model to output a continuous physical condition related to a specific disease as a numerical value using the electrocardiogram data. The neural network model can be the neural network model described above with reference to FIGS. 2 and 3.

[0079] The computing device 100 can train the neural network model based on at least one of a first feature for biological information indicating a physical characteristic correlated with a disease or a second feature for pathological information reflecting the degree of progression of the disease.

[0080] The computing device 100 can train the neural network model by supervised learning or self-supervised learning. Specifically, the computing device 100 can train the neural network model by inputting various disease-related factors and numerical values ​​indicating the association between each factor and the disease in electrocardiogram data as training data to the neural network model. Alternatively, the computing device 100 can train the neural network model using training data including unlabeled electrocardiogram data so that the neural network model extracts disease-related factors on its own and outputs numerical values ​​indicating the association between each factor and the disease.

[0081] The pre-trained neural network model can receive electrocardiogram data and output a disease-related value, which indicates the physical condition of the subject whose electrocardiogram data was measured and is a disease-related value, and can indicate whether the disease has occurred or the progression of the disease.

[0082] FIG. 5 is a flowchart illustrating an inference method for sub-models constituting a neural network model according to an embodiment of the present disclosure.

[0083] 5, the computing device 100 according to an embodiment of the present disclosure may acquire electrocardiogram data (S210), which is similar to step S110 of FIG. 4 and will not be described in detail.

[0084] The computing device 100 may output a first feature for the biological information through a first sub-model (S220). The first sub-model may be trained to output the first feature based on electrocardiogram data. The first feature may include at least one factor included in the biological information or a numerical value for the factor. There may be multiple first sub-models, and each of the first sub-models may individually output at least one factor included in the biological information. Thus, the number of first sub-models may be configured according to the number of factors.

[0085] The computing device 100 may output second features for the pathological information through the second sub-model (S230). The second sub-model may be trained to output the second features based on the electrocardiogram data. The second features may include at least one factor included in the pathological information or a numerical value for the factor. There may be multiple second sub-models, and each second sub-model may individually output at least one factor included in the pathological information. Therefore, the number of second sub-models may be configured according to the number of factors.

[0086] Here, each of the first sub-model and the second sub-model may be trained based on self-supervised learning performed using training data including samples for which labels have not been assigned.

[0087] The computing device 100 may output a numerical value for the physical condition through a third sub-model based on the first feature output from the first sub-model and the second feature output from the second sub-model (S240). The third sub-model may be trained to numerically represent the physical condition, which continuously changes depending on the onset or progression of a disease, based on the first feature and the second feature. The computing device 100 may determine a weight value for combining the first feature and the second feature according to the type of disease. The computing device 100 may combine the first feature and the second feature according to the determined weight value to generate a third feature. Thus, the third sub-model may output a numerical value indicating the physical condition based on the third feature.

[0088] Meanwhile, although steps S220 and S230 are shown to be performed sequentially in FIG. 5, steps S220 and S230 may be performed in parallel.

[0089] The various embodiments of the present disclosure described above can be combined with additional embodiments and can be modified within the scope that can be understood by those skilled in the art from the above detailed description. The embodiments of the present disclosure are illustrative in all respects and should not be construed as limiting. For example, each component described as a single type can also be implemented in a distributed form, and similarly, components described as distributed can also be implemented in a combined form. Therefore, all modifications and variations derived from the meaning, scope, and equivalent concepts of the claims of the present disclosure should be construed as being within the scope of the present disclosure. [Explanation of symbols]

[0090] 100: Computing equipment 110: Processor 120: Memory 130: Network Department 200: Neural network model 210: First submodel 211: First feature 220: Second submodel 221: Second feature 230: 3rd submodel 300: Electrocardiogram data 400: Physical condition

Claims

1. 1. A method for measuring continuous physical states based on deep learning, performed by a computing device including at least one processor, comprising: acquiring electrocardiogram data; using a pre-trained neural network model to infer, based on the electrocardiogram data, a physical state corresponding to the onset of a disease or the progression of the disease in the subject for whom the electrocardiogram data was measured; Including, The method, wherein the neural network model is trained based on at least one of a first feature for biological information indicating physical characteristics correlated with the disease, or a second feature for pathological information reflecting the degree of progression of the disease.

2. the neural network model includes a first sub-model trained to output the first feature based on the electrocardiogram data; The method of claim 1 , wherein the first sub-model is configured to individually output a numerical value for at least one factor included in the biological information in accordance with the number of the factors.

3. the neural network model further includes a second sub-model trained to output the second feature based on the electrocardiogram data; The method according to claim 2 , wherein the second sub-model is configured to individually output a numerical value for at least one factor included in the pathological information in accordance with the number of the factors.

4. The method according to claim 3, wherein the neural network model further includes a third sub-model trained to numerically represent a physical condition that continuously changes depending on the onset or progression of the disease, based on a first feature that is the output of the first sub-model and a second feature that is the output of the second sub-model.

5. The method according to claim 4 , wherein the third sub-model receives a third feature generated by combining the first feature and the second feature based on a weight determined by the type of the disease and outputs the numerical value.

6. The method of claim 3 , wherein each of the first sub-model and the second sub-model is trained based on self-supervised learning performed using training data including samples for which labels were not assigned.

7. The method of claim 1 , wherein the disease comprises cardiovascular disease.

8. The method according to claim 7 , wherein the biological information includes at least one of age, sex, height, and weight as a physical characteristic factor related to coronary artery disease included in the cardiovascular disease.

9. The method according to claim 8, wherein the pathological information includes at least one of the following pathological characteristic factors reflecting the degree of progression of coronary artery disease included in the cardiovascular disease: presence or absence of myocardial infarction, degree of vascular calcification, thrombus safety, coronary artery intravascular velocity, or degree of coronary artery stenosis.

10. A computer program stored on a computer-readable storage medium, the computer program performing an operation of measuring continuous physical conditions based on deep learning when executed by one or more processors; The operation is acquiring electrocardiogram data; an operation of inferring a physical state corresponding to the onset of a disease or the progression of a disease in a subject whose electrocardiogram data was measured based on the electrocardiogram data, using a pre-trained neural network model; Including, A computer program, wherein the neural network model is trained based on at least one of a first feature for biological information indicating a physical characteristic correlated with the disease, or a second feature for pathological information reflecting the degree of progression of the disease.

11. A computing device for measuring continuous physical states based on deep learning, comprising: a processor including at least one core; a memory containing program code executable by the processor; a network unit for acquiring electrocardiogram data; Including, The processor uses a neural network model trained based on at least one of a first feature for biological information indicating a physical characteristic of the subject correlated with a disease or a second feature for pathological information reflecting the degree of progression of the disease to infer a physical condition corresponding to the onset of a disease or the progression of the disease in the subject for which the electrocardiogram data was measured, based on the electrocardiogram data.

Images