Method, device, and computer program for constructing neural network model suitable for user environment

Abstract

The present disclosure provides a method, a device, and a computer program for constructing a neural network model suitable for a user environment. The method according to one embodiment of the present disclosure comprises the steps of: inputting acquired biological data of a user into a pre-trained first neural network model to acquire a first score corresponding to a possibility of disease; and predicting a state of the user on the basis of the first score, wherein the model may be selected to correspond to the type of a medical institution having acquired the biological data of the user, from among a plurality of second neural network models trained to predict the state on the basis of a plurality of pieces of biological data of the same type as the biological data.

Description

Methods, devices and computer programs for building neural network models tailored to user environments

[0001] The present disclosure relates to a method, device, and computer program for building a neural network model tailored to a user's environment. Specifically, the present disclosure relates to a method, device, and computer program for building a neural network model that identifies the environment in which a user's biometric data is acquired and derives more accurate analysis results for the biometric data acquired in that environment.

[0002] Recent advancements in information and communication technology have led to the application of artificial intelligence (AI) in various fields. In particular, AI is being utilized in various fields, including the medical field, where previously relied on specialized and limited resources such as doctors and researchers to diagnose patients' illnesses. For example, this involves inputting biometric data acquired in real time from a patient into a pre-trained AI model to analyze the patient's condition or diagnose a disease.

[0003] Furthermore, while biosignal analysis, such as electrocardiogram measurements, was previously performed in hospitals using specialized medical equipment, electrocardiogram measurements are now possible using portable electronic devices such as smartwatches. Consequently, deep learning technology allows users to analyze electrocardiogram data without expert intervention, automatically diagnosing and predicting conditions like myocardial infarction.

[0004] However, AI models used in these medical fields are trained primarily on data from patients visiting hospitals. Patients receiving electrocardiograms at hospitals often have underlying conditions or symptoms, and a high proportion are middle-aged or older. Therefore, AI models built on this data primarily rely on ECG data from hospital patients, and are limited in their ability to sufficiently learn from data from relatively younger age groups, who typically use smartwatches. Consequently, if AI models trained on existing hospital data are directly applied to smartwatch users, a mismatch between the training data and the data used in the actual inference process may arise.

[0005] The present disclosure has been made in response to the aforementioned background technology, and aims to provide a method, device, and computer program for constructing a neural network model suitable for a user environment.

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

[0007] In accordance with an embodiment of the present disclosure for realizing the task described above, a method for constructing a neural network model suitable for a user environment, performed by a computing device including at least one processor, comprises the steps of: inputting acquired user's biometric data into a pre-trained first neural network model to obtain a first score corresponding to a possibility of a disease; and predicting the user's condition based on the first score, wherein the first neural network model is selected from among a plurality of neural network models according to the type of medical institution that acquired the user's biometric data.

[0008] Alternatively, the step of predicting the user's condition includes a step of predicting that the user is included in a high-risk group for the disease if the first score is greater than or equal to a first reference value, wherein the first reference value is set corresponding to a type of location where the user's biometric data was acquired.

[0009] Alternatively, the method includes a step of assigning a first label value related to the possibility of the disease to the acquired user's biometric data and setting it as learning data if the first score is equal to or higher than a preset value of the first reference value, and transmitting the learning data to another device that uses the same neural network model as the first neural network model.

[0010] Alternatively, the method includes a step of performing primary learning based on a first learning data set including a plurality of biometric data acquired from a plurality of users to obtain a foundation model, a step of constructing a plurality of second learning data sets corresponding to the types of the plurality of medical institutions based on biometric data acquired from at least one medical institution corresponding to each type of the medical institution, and a step of performing secondary learning on the previously learned foundation model based on the plurality of second learning data sets to obtain the plurality of neural network models.

[0011] Alternatively, the secondary learning comprises transfer learning performed based on the second learning data set.

[0012] Alternatively, the method includes a step of selecting a second neural network model corresponding to the new type of medical institution among the plurality of neural network models when it is detected that the type of medical institution that acquires the user's biometric data has changed, and inputting the acquired user's biometric data into the pre-trained second neural network model to acquire a first score corresponding to the possibility of a disease.

[0013] Alternatively, the method includes identifying a location of the user at the time the biometric data was acquired through the sensor, and determining the type of medical institution that acquired the biometric data of the user based on the identified location.

[0014] Alternatively, the plurality of neural network models are characterized by including a neural network model corresponding to a tertiary hospital, a neural network model corresponding to a secondary hospital, a neural network model corresponding to a primary hospital, a neural network model corresponding to a nursing hospital, a neural network model corresponding to a health checkup center, and a neural network model corresponding to a non-medical institution.

[0015] Alternatively, if the computing device is a smartwatch, the first neural network model is characterized in that it is selected as a neural network model corresponding to a non-medical institution.

[0016] A computer program stored in a computer-readable storage medium according to one embodiment of the present disclosure for realizing the task as described above, wherein the computer program, when executed on one or more processors, performs an operation of constructing a neural network model suitable for a user environment, the operation including an operation of inputting acquired user's biometric data into a pre-trained first neural network model to obtain a first score corresponding to a possibility of a disease, and an operation of predicting the user's condition based on the first score, wherein the first neural network model is characterized in that it is selected from among a plurality of neural network models according to the type of medical institution that obtained the user's biometric data.

[0017] In order to realize the task as described above, a computing device for constructing a neural network model suitable for a user environment according to an embodiment of the present disclosure comprises a communication interface for acquiring a user's biometric data, a memory including a program code, and at least one processor for inputting the acquired user's biometric data into a pre-trained first neural network model to acquire a first score corresponding to a possibility of a disease, and predicting the user's condition based on the first score, wherein the first neural network model is selected from among a plurality of neural network models according to the type of medical institution that acquired the user's biometric data.

[0018] According to one embodiment of the present disclosure, a method for building a neural network model tailored to a user environment can effectively combine data collected in a hospital setting with data measured via a smartwatch to build an AI disease prediction model optimized for the user environment. This overcomes the limitations of models trained primarily on patients visiting existing hospitals, enabling reliable disease prediction for smartwatch users. Furthermore, by resolving data inconsistency issues, the AI ​​model's prediction accuracy is improved, and real-time personalized health monitoring and disease prediction can be provided without expert intervention.

[0019] FIG. 1 is an exemplary diagram of a method for obtaining a neural network model suitable for a user environment according to an embodiment of the present disclosure.

[0020] FIG. 2 is a schematic block diagram of a computing device according to one embodiment of the present disclosure.

[0021] FIG. 3 is a flowchart of a method for managing a user's health status based on user context information according to an embodiment of the present disclosure.

[0022] FIG. 4 is an exemplary diagram illustrating a second neural network model according to a type of context information according to an embodiment of the present disclosure.

[0023] FIG. 5 is an exemplary diagram illustrating a method for determining a correlation between dietary information and disease potential according to one embodiment of the present disclosure.

[0024] FIG. 6 is an exemplary diagram illustrating a method for determining a correlation between exercise information and disease possibility according to one embodiment of the present disclosure.

[0025] FIG. 7 is an exemplary diagram illustrating a method for determining a correlation between medication information and disease possibility according to one embodiment of the present disclosure.

[0026] FIG. 8 is an exemplary diagram illustrating the preparation of multiple first neural network models according to an organ that acquired biometric data according to one embodiment of the present disclosure.

[0027] FIG. 9 is a block diagram of a computing device according to another embodiment of the present disclosure.

[0028] Below, embodiments of the present disclosure are described in detail with reference to the attached drawings so that those skilled in the art can easily implement the present disclosure. The embodiments presented in this disclosure are provided to enable those skilled in the art to utilize or implement the contents of the present disclosure. Accordingly, various modifications to the embodiments of the present disclosure will be apparent to those skilled in the art. That is, the present disclosure may be implemented in various different forms and is not limited to the embodiments described below.

[0029] Throughout the specification of this disclosure, identical or similar drawing numbers refer to identical or similar components. Furthermore, for clarity in the description of this disclosure, drawing numbers for parts unrelated to the description of this disclosure may be omitted in the drawings.

[0030] The term "or" as used herein is intended to mean an inclusive "or" rather than an exclusive "or." That is, unless otherwise specified herein or clear from context, "X employs A or B" should be understood to mean either of the natural inclusive permutations. For example, unless otherwise specified herein or clear from context, "X employs A or B" can be interpreted to mean either X employs A, X employs B, or X employs both A and B.

[0031] The term "and / or" as used herein should be understood to refer to and include all possible combinations of one or more of the related concepts listed.

[0032] The terms "comprises" and / or "comprising" as used herein should be understood to mean the presence of certain features and / or components. However, it should be understood that the terms "comprises" and / or "comprising" do not exclude the presence or addition of one or more other features, other components, and / or combinations thereof.

[0033] Unless otherwise specified in this disclosure or unless the context makes it clear that the singular form is intended to be referred to, the singular should generally be construed to include “one or more.”

[0034] The term "Nth (N is a natural number)" used in this disclosure can be understood as an expression used to mutually distinguish components of this disclosure based on a predetermined standard such as a functional perspective, a structural perspective, or convenience of explanation. For example, components performing different functional roles in this disclosure can be distinguished as a first component or a second component. However, components that are substantially the same within the technical spirit of this disclosure but must be distinguished for convenience of explanation may also be distinguished as a first component or a second component.

[0035] The term "acquisition" as used in this disclosure may be understood to mean not only receiving data through a wired or wireless communication interface with an external device or system, but also generating data in an on-device form.

[0036] Meanwhile, the term "module" or "unit" used in the present disclosure can be understood as a term referring to an independent functional unit that processes computing resources, such as a computer-related entity, firmware, software or a part thereof, hardware or a part thereof, or a combination of software and hardware. At this time, the "module" or "unit" may be a unit composed of a single element, or a unit expressed as a combination or set of multiple elements. For example, as a narrow concept, a "module" or "unit" may refer to a hardware element of a computing device or a set thereof, an application program that performs a specific function of software, a processing process implemented through software execution, or a set of instructions for program execution, etc. In addition, as a broad concept, a "module" or "unit" may refer to the computing device itself that constitutes the system, or an application running on the computing device, etc. However, since the above-described concept is only an example, the concept of “module” or “part” may be defined in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

[0037] The term "model" as used herein may be understood as a system implemented using mathematical concepts and language to solve a specific problem, a set of software units to solve a specific problem, or an abstract model of a processing process to solve a specific problem. For example, a neural network "model" may refer to the entire system implemented as a neural network that has problem-solving capabilities through learning. In this case, the neural network can have 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 a combination of multiple neural networks.

[0038] The term "data" used in this disclosure may include "images," signals, and the like. The term "image" used in this disclosure may refer to multidimensional data composed of discrete image elements. In other words, "image" may be understood as a term referring to a digital representation of an object visible to the human eye. For example, "image" may refer to multidimensional data composed of elements corresponding to pixels in a two-dimensional image. "Image" may refer to multidimensional data composed of elements corresponding to voxels in a three-dimensional image.

[0039] The explanation of the above terms is intended to aid understanding of the present disclosure. Therefore, unless explicitly stated as limiting the contents of the present disclosure, it should be noted that the above terms are not intended to limit the technical ideas of the contents of the present disclosure.

[0040] FIG. 1 is an exemplary diagram of a method for managing a user's health status based on user context information according to one embodiment of the present disclosure.

[0041] A computing device (100) according to an embodiment of the present disclosure may be a hardware device or a part of a hardware device that performs comprehensive processing and calculation of data, or may be a software-based computing environment connected to a communication interface. For example, the computing device (100) may be a server that performs intensive data processing functions and shares resources, or may be a client that shares resources through interaction with a server. In addition, the computing device (100) may be a cloud system in which multiple servers and clients interact to comprehensively process data. Since the above description is only one example related to the type of computing device (100), the type of computing device (100) may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

[0042] Referring to FIG. 1, a computing device (100) according to an embodiment of the present disclosure may be implemented as a server device that is linked to an external electronic device (200) (i.e., a smart watch) worn by a user (1). However, the present disclosure is not limited thereto, and the computing device (100) may be implemented as various electronic devices such as a desktop, a laptop, a smartphone, a smart watch, a smart band, a smart ring, etc. In the following, for the purpose of understanding the present disclosure, the computing device (100) will be described as a server device.

[0043] A computing device (100) according to an embodiment of the present disclosure can obtain biometric data of a user (1). For example, the computing device (100) can obtain electrocardiogram data as biometric data of the user (1). Specifically, the electrocardiogram data may be data obtained by digitizing electrocardiogram data obtained by detecting an electrical signal generated from the heart of the user (1) through a smartwatch. However, the present invention is not limited thereto, and the computing device (100) can obtain various biometric data from the user (1) in addition to electrocardiogram data. For example, the computing device (100) can obtain various biometric data such as pulse, body temperature, blood pressure, blood flow, electroencephalogram signal, etc. of the user (1) by using various biometric signal measuring devices (or sensors, etc.).

[0044] According to one embodiment of the present disclosure, a computing device (100) may store a pre-trained neural network model (10). The neural network model (10) may be a model (10) trained to predict the condition of a user (1) based on electrocardiogram data of the user (1). The neural network model (1) may be a model (10) trained to predict the condition of the user (1) by extracting feature information related to the condition of the user (1) inherent in the electrocardiogram data of the user (1). The condition of the user (1) may be a condition related to the health of the user (1), and in particular, may be related to a disease. That is, the neural network model (10) may be a model (10) trained to predict a specific disease potential in the user (1) based on the electrocardiogram data. Here, the disease may be a heart disease. For example, heart disease may include left ventricular systolic dysfunction (LVSD), myocardial infarction, heart failure, arrhythmia, coronary artery disease (CAD), and cardiomyopathy.

[0045] The neural network model (10) can be trained in advance to predict heart disease based on the electrocardiogram data input as learning data, which includes a plurality of electrocardiogram data and heart disease information (e.g., type of heart disease, presence or absence of heart disease, degree of heart disease, etc.) matching each of the electrocardiogram data. At this time, the plurality of electrocardiogram data included in the learning data may each include feature information associated with heart disease (e.g., feature information included in the electrocardiogram data, such as P wave, Q wave, R wave, QT interval, QRS width, T wave, etc. observed in the electrocardiogram data). Through this, when the electrocardiogram data is input, the neural network model (10) can be trained to extract feature information corresponding to heart disease from the electrocardiogram data and predict the possibility of heart disease based on the extracted feature information. In particular, the neural network model (10) can be trained in advance to calculate the possibility of heart disease (e.g., probability value) as a score. At this time, the neural network model (10) may include a plurality of sub-neural network models classified according to the type of disease to be predicted.

[0046] In addition, the pre-learned neural network model (10) may include a plurality of neural network models (10) learned to predict a potential disease of a user (1) based on various bio-information obtainable from a computing device (100) as well as electrocardiogram data. In this case, each neural network model (10) may be learned in advance as learning data disease information (e.g., disease presence / absence information, etc.) to be predicted or identified based on bio-information and bio-information matching the bio-information.

[0047] The neural network model (10) can be implemented as a multi-layer perceptron (MLP), a convolutional neural network (CNN), a recurrent neural network (RNN), a generative adversarial network, etc. In particular, the neural network model (10) can include a plurality of residual blocks that extract latent feature information of input electrocardiogram data and a classifier that classifies the presence or absence of a disease based on the extracted latent feature information.

[0048] Meanwhile, referring to FIG. 1, the computing device (100) may include multiple neural network models corresponding to the environment of the user (1) from whom biometric data is acquired. In this regard, a neural network model trained to predict a specific disease based on acquired biometric data, as illustrated in FIG. 1, may be trained based on biometric data acquired from patients visiting a hospital. That is, the training data set of the neural network model may consist of biometric data acquired from patients visiting a hospital. Subjects whose electrocardiograms are measured at hospitals often have underlying diseases or symptoms, and a high proportion of those in the middle-aged or elderly may be present. Therefore, a neural network model built based on such biometric data primarily trains on biometric data from patients visiting the hospital, and has limitations in that it cannot sufficiently train on biometric data acquired from smartwatches, i.e., biometric data from relatively young age groups who primarily use smartwatches. Accordingly, if a neural network model trained primarily on existing hospital data is directly applied to users who use smartwatches, a mismatch between the training data and the biometric data that must perform the actual inference process may arise. Accordingly, a computing device according to an embodiment of the present disclosure may include a plurality of neural network models corresponding to a user environment in which biometric data is acquired. Here, the plurality of neural network models may correspond to the type of environment of the user (1) in which preset biometric data is acquired. For example, the computing device (100) may include a plurality of neural network models corresponding to the type of medical institution in which biometric data is acquired, such as a tertiary hospital, a secondary hospital, etc.

[0049] At this time, multiple neural network models can be trained based on biometric data acquired in each user environment. That is, a neural network model corresponding to a smartwatch can be trained based on a training data set comprised of biometric data measured from the smartwatch. Conversely, a neural network model corresponding to a university hospital, such as a tertiary hospital, can be trained based on a training data set comprised of biometric data acquired from the university hospital. This allows for the application of neural network models optimized for each user environment, thereby overcoming the limitations of models trained with existing hospital-centric data and enabling more accurate disease prediction in diverse environments. Furthermore, by considering differences in the environments in which biometric data is acquired and selecting and utilizing neural network models, the problem of discrepancies between the training data and actual inference data can be minimized. Consequently, the computing device according to the present disclosure provides a user-tailored disease prediction model, enabling more reliable health monitoring.

[0050] Hereinafter, an embodiment of the present disclosure will be described in detail with reference to FIGS. 2 to 8.

[0051] FIG. 2 is a schematic block diagram of a computing device (100) according to one embodiment of the present disclosure.

[0052] Referring to FIG. 2, a computing device (100) according to an embodiment of the present disclosure may include one or more processors (hereinafter, “processors”) (110), a memory (120), and a communication interface (130). However, FIG. 2 is merely an example, and thus the computing device (100) may further include other components for implementing a computing environment. In addition, only some of the disclosed components may be included in the computing device (100).

[0053] In order to increase the accuracy of predicting the likelihood of a disease of a user (1) based on electrocardiogram data, a computing device (100) according to an embodiment of the present disclosure can periodically acquire electrocardiogram data of a user (1) to predict the likelihood of a disease. To this end, the computing device (100) can periodically request a smartwatch to measure an electrocardiogram signal, or can periodically acquire electrocardiogram data of the user (1) from the smartwatch according to the user's (1) settings. If the computing device (100) determines that the likelihood of a disease of the user (1) is high based on the acquired electrocardiogram data, the computing device (100) can identify the cause of the likelihood of the disease based on the context information of the user (1). In particular, if the trend of the score corresponding to the likelihood of a disease suddenly changes and it is determined that the likelihood of a disease of the user (1) is high, the computing device (100) can analyze the context information of the user (1) to identify the cause. Here, the context information includes various information that can be acquired in relation to the status, environment, and activities of the user (1). For example, context information may include information on nutrients consumed by the user (1), information on medication taken by the user (1), information on restaurants visited by the user (1), information on rest and sleep time of the user (1), exercise information of the user (1), etc. The computing device (100) may determine whether the cause of the user's (1) sudden high likelihood of illness, i.e., the cause of the high score, is related to the context information. That is, the computing device (100) may determine the relationship between the likelihood of illness and the context information. At this time, if the computing device (100) determines that the cause of the user's (1) sudden high likelihood of illness is related to the context information, the computing device (100) may provide guide information related to the context information. The guide information may be information suggesting adjusting or resetting the user's (1) status, environment, activity, etc.If the computing device (100) determines that the cause of the user's (1) sudden high likelihood of illness is not related to context information, the computing device (100) may provide the user (1) with information suggesting that he or she visit a medical facility such as a hospital or public health center.

[0054] A processor (110) according to an embodiment of the present disclosure may be understood as a configuration unit including hardware and / or software for performing computing operations. For example, the processor (110) may read a computer program to perform data processing for machine learning. The processor (110) may process computational processes such as processing input data 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), or a field programmable gate array (FPGA). The above-described type of processor (110) is only one example, and thus, the type of processor (110) may be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

[0055] The processor (110) is electrically connected to other components of the computing device (100) (i.e., memory (120) and communication interface (130)) and controls the overall operation of the computing device (100).

[0056] The memory (120) according to one embodiment of the present disclosure may be understood as a configuration unit including hardware and / or software for storing and managing data processed in the computing device (100). That is, the memory (120) may store any type of data generated or determined by the processor (110) and any type of data received by the communication interface. For example, the memory (120) may include at least one type of storage medium among a flash memory type, a hard disk type, a multimedia card micro type, a card type memory, a random access memory (RAM), a static random access memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, and an optical disk. In addition, the memory (120) may also include a database system that controls and manages data in a predetermined system. The type of memory (120) described above is only one example, and thus the type of memory (120) can be configured in various ways within a range understandable to those skilled in the art based on the contents of the present disclosure.

[0057] The memory (120) can structure and organize and manage data, combinations of data, and program codes executable by the processor (110) required for the processor (110) to perform operations. For example, the memory (120) may store a neural network model (10) trained to predict the possibility of a disease by inputting biometric data, and a neural network model trained to identify the correlation between context information and the possibility of a disease. In addition, the memory (120) may include program codes and learning data (e.g., a plurality of electrocardiogram signals and a plurality of heart disease information matching each of the electrocardiogram signals) that operate to perform learning for the above-described neural network model. In addition, the memory (120) may store program codes that operate the neural network model (10) to receive an electrocardiogram signal and perform inference according to the purpose of use of the computing device (100), and processed data generated as the program codes are executed.

[0058] The communication interface (130) can be understood as a component that transmits and receives data through any type of known wired or wireless communication system. For example, the communication interface (130) can perform data transmission and reception using a wired or wireless communication system such as a local area network (LAN), wideband code division multiple access (WCDMA), long term evolution (LTE), wireless broadband internet (WiBro), 5th generation mobile communication (5G), ultrawide-band, ZigBee, radio frequency (RF) communication, wireless LAN, wireless fidelity, near field communication (NFC), or Bluetooth. Since the above-described communication systems are only examples, the wired and wireless communication system for data transmission and reception of the communication interface (130) can be applied in various ways other than the above-described examples. The communication interface (130) can receive data necessary for the processor (110) to perform calculations through wired or wireless communication with any system or any client, etc. In addition, the communication interface (130) can transmit data generated through calculations of the processor (110) through wired or wireless communication with any system or any client, etc. For example, the communication interface (130) can receive medical data through communication with a cloud server that performs tasks such as standardization of databases and medical data within a hospital environment, or a computing device, etc.The communication interface (130) can transmit output data of the neural network model (10), intermediate data, processed data, etc. derived from the computational process of the processor (110), etc. through communication with the aforementioned database, server, or computing device. For example, the processor (110) can obtain biometric data from an external computing device (e.g., a smart watch) through the communication interface.

[0059] FIG. 3 is a flowchart of a method for managing the health status of a user (1) based on user (1) context information according to one embodiment of the present disclosure.

[0060] Referring to FIG. 3, the processor (110) inputs the acquired biometric data of the user (1) into a pre-learned first neural network model to obtain a first score corresponding to the possibility of a disease (S310).

[0061] Specifically, the processor (110) can obtain biometric data of the user (1) from a biometric signal measuring device (e.g., a smartwatch worn by the user (1) in FIG. 1) that is linked to a computing device. Then, the processor (110) can input the obtained biometric data into a pre-trained neural network model (hereinafter, the first neural network model) to obtain a score (hereinafter, the first score) corresponding to the possibility of a disease. In the following, for the convenience of explanation of the present disclosure, the biometric data will be explained as electrocardiogram data and the disease as left ventricular systolic dysfunction.

[0062] According to one embodiment of the present disclosure, when electrocardiogram data is input, the first neural network model may be pre-trained to extract feature information from the electrocardiogram data and produce a score corresponding to the possibility of left ventricular systolic dysfunction. Specifically, the pre-trained neural network model (10) may be trained based on input data including a plurality of electrocardiogram data and training data including label data in which a label indicating the presence or absence of left ventricular systolic dysfunction (or a numerical value of the degree) for each input data is assigned, and may be trained to minimize an error between the output corresponding to the input data and the label data by a backpropagation algorithm. When electrocardiogram data is input, the trained neural network model (10) may extract feature information (e.g., feature points observed on the electrocardiogram signal) from the electrocardiogram data and output a score (or probability value) regarding the possibility of left ventricular systolic dysfunction of the subject of the examination.

[0063] In particular, the processor (110) can periodically acquire electrocardiogram data to obtain a first score corresponding to the possibility of left ventricular contractile function of the user (1) and monitor the condition of the user (1).

[0064] At this time, according to one embodiment of the present disclosure, if the first score is equal to or greater than a first reference value, the processor (110) may determine that there is a high possibility that the user (1) has a disease, acquire context information of the user (1), and determine a correlation between the acquired context information and the disease. Specifically, if the first score is equal to or greater than a reference value (hereinafter, referred to as the first reference value), the processor (110) may determine that there is a high possibility that the user (1) has left ventricular systolic dysfunction. At this time, the reference value may include a plurality of reference values ​​corresponding to the plurality of grades by classifying the possibility of left ventricular systolic dysfunction into a plurality of grades. For example, a 1-1 reference value that determines a normal state, a 1-2 reference value corresponding to a caution grade, a 1-3 reference value corresponding to a medium risk group, and a 1-4 reference value corresponding to a high risk group may be set, and the first reference value may be the 1-4 reference value corresponding to the high risk group. When the processor (110) determines that the user (1) is at high risk for left ventricular systolic dysfunction, it can obtain context information from the user (1) and determine the relationship between the obtained context information and left ventricular systolic dysfunction. Meanwhile, multiple grades set in relation to the disease (i.e., left ventricular systolic dysfunction) can be set in various forms and numbers.

[0065] The processor (110) monitors the trend of the first score, and if the user (1) is suddenly determined to have a high possibility of left ventricular systolic dysfunction (or is determined to be in a high-risk group for left ventricular systolic dysfunction), context information can be obtained from the user (1). For example, if the error between the average value of a plurality of previously obtained first scores and the last obtained first score exceeds a preset value, and the last obtained first score is equal to or greater than a first reference value, context information can be obtained from the user (1), and the correlation between the obtained context information and left ventricular systolic dysfunction can be determined. That is, if the processor (110) determines that the condition of the user (1) related to a disease has suddenly changed, the processor (110) can determine the correlation between the context information of the user (1) and the disease, and analyze the cause thereof.

[0066] Meanwhile, the processor (110) can periodically acquire context information, similar to electrocardiogram data. At this time, if the first score is greater than or equal to the first reference value, the processor (110) can select context information close to the time point at which the electrocardiogram data corresponding to the first score calculated to be greater than or equal to the first reference value was acquired, and determine the correlation between the selected context information and the disease.

[0067] The processor (110) can determine the correlation between context information and a disease. Determining the correlation may involve determining the extent to which the context information affects a first score corresponding to the likelihood of a disease. To this end, according to one embodiment of the present disclosure, the processor (110) inputs context data corresponding to the context information of the user (1) and the acquired biometric data of the user (1) into a pre-trained neural network model (hereinafter, “second neural network model”) to obtain a score corresponding to the correlation (hereinafter, “second score”), and based on the acquired second score, the correlation between the acquired context information and the disease can be determined.

[0068] Specifically, the second neural network model may be a model trained to output a second score indicating the degree of correlation between the disease possibility and the context information by inputting context data and electrocardiogram data corresponding to the context information. The context data may be configured as a one-dimensional matrix with values ​​that quantify the type of context information, the status, environment, and activity of the user (1) included in the context information. For example, if the context information is the user's (1) diet information, the context data may be configured with a sodium level calculated by analyzing the diet information. Alternatively, if the context information is the user's (1) exercise information, the context data may be configured with an exercise type, a value set for exercise intensity corresponding to the exercise type, and an exercise time. Alternatively, if the context information is the user's (1) medication information, the context data may be configured with a value set corresponding to the medication taken by the user (1) identified by analyzing the user's (1) medication information, and the number of medications taken. Alternatively, if the context information is the user's (1) status information, the context data may be configured with the user's (1) rest time or sleep time identified by analyzing the user's (1) status.

[0069] The second neural network model may be a model trained to identify a correlation between multiple electrocardiogram data and context data corresponding to the multiple electrocardiogram data. In particular, the second neural network model may be a model trained to take electrocardiogram data and context data as input and produce a score corresponding to the correlation. Specifically, the second neural network model may be trained based on training data including multiple electrocardiogram data and context data corresponding to the multiple electrocardiogram data. For example, the second neural network model may be trained based on training data including electrocardiogram data and sodium levels (or scores corresponding to sodium levels) measured from the same patient as input data, and including a correlation score between the electrocardiogram data and sodium levels as label data. The second neural network model may be trained to minimize an error between an output corresponding to the input data and the label data by a backpropagation algorithm. When electrocardiogram data and context data are input, the neural network model learned in this way can extract feature information (e.g., feature points observed on the electrocardiogram signal) from the electrocardiogram data, determine the correlation between the extracted feature information and the context data, and output a score (or probability value) regarding the possibility of the subject of the examination having left ventricular systolic dysfunction. The processor (110) can determine that the higher the second score, the higher the correlation between the context data and the electrocardiogram data. To this end, the second neural network model can be implemented as a multi-layer perceptron (MLP), a convolutional neural network (CNN), a recurrent neural network (RNN), a generative adversarial network, etc.

[0070] FIG. 4 is an exemplary diagram illustrating a second neural network model according to a type of context information according to an embodiment of the present disclosure.

[0071] Meanwhile, referring to FIG. 4, context information acquired from a user (1) may be of various types, and thus the second neural network model may include a plurality of sub-neural network models corresponding to a plurality of types of context information. Accordingly, the processor (110) selects a sub-neural network model corresponding to the type of acquired context information, and inputs context data and biometric data corresponding to the acquired context information into the selected sub-neural network model, thereby identifying a correlation between the context information and a disease.

[0072] According to one embodiment of the present disclosure, context information of a user (1) may include dietary information of the user (1). In this case, the processor (110) may determine a correlation between the diet of the user (1) and the possibility of the user (1) developing a disease.

[0073] Specifically, if the processor (110) determines that the user (1) has a high possibility of having a disease (or if the first score is identified as being equal to or greater than the first reference value), the processor (110) may obtain the user's (1) diet information. For example, the processor (110) may request a camera (specifically, a camera included in a computing device or a camera included in an external electronic device (e.g., a smart phone, a smart watch, etc.) worn by the user (1)) to take a picture of the food consumed by the user (1), thereby obtaining the food image. Alternatively, the processor (110) may input information on the food consumed by the user (1). Then, the processor (110) may analyze the diet information to determine whether the user's (1) diet affects the possibility of having a disease or the extent to which it affects the user's (1) diet.

[0074] FIG. 5 is an exemplary diagram illustrating a method for determining a correlation between dietary information and disease potential according to one embodiment of the present disclosure.

[0075] In particular, the processor (110) can obtain images of food consumed by the user (1) as the user's (1) diet information, analyze the images of the consumed food, determine the sodium level consumed by the user (1), and determine the correlation between the determined sodium level and the possibility of the user's (1) disease. Specifically, referring to FIG. 5, the processor (110) can determine the sodium level consumed by the user (1) by inputting the food image into a pre-trained neural network model (the third neural network model). At this time, the third neural network model can be trained to extract food included in the food image, determine the type of food, and calculate the sodium level consumed by the user (1). The third neural network model can be trained based on training data composed of input data including a plurality of food images and label data in which segmentation of food included in the food image and a label for the type of food are assigned. The third neural network model can also be implemented using a multi-layer perceptron (MLP), a convolutional neural network (CNN), a recurrent neural network (RNN), a generative adversarial network, etc. In addition, the processor (110) inputs sodium levels and electrocardiogram data into the second neural network model to obtain a second score and determine the correlation between sodium levels and diseases.

[0076] FIG. 6 is an exemplary diagram illustrating a method for determining a correlation between exercise information and disease possibility according to one embodiment of the present disclosure.

[0077] Additionally, according to one embodiment of the present disclosure, context information of the user (1) may include exercise information of the user (1). The processor (110) may determine a correlation between exercise performed by the user (1) and the possibility of disease of the user (1) based on the exercise type, exercise intensity, and exercise time included in the exercise information of the user (1).

[0078] Referring to FIG. 6, if the processor (110) determines that the user (1) has a high possibility of being ill (or if the first score is identified as being equal to or greater than a first reference value), the processor (110) may obtain exercise information of the user (1). For example, the processor (110) may detect whether the user (1) is exercising through a sensor included in an external electronic device (e.g., a smart phone, a smart watch, etc.) worn by the user (1). If the processor (110) determines that the user (1) is in a gym or fitness center through a GPS sensor and determines that the user (1) is moving based on a gyro sensor, the processor (110) may determine that the user (1) is exercising. Alternatively, the processor (110) may receive information on exercise performed by the user (1). Then, the processor (110) may analyze the obtained exercise information to determine the type of exercise performed by the user (1), the exercise time, and the exercise intensity, and generate context data. The exercise intensity may be determined based on the exercise type and the exercise time. In addition, the processor (110) inputs context data and electrocardiogram data into a pre-trained second neural network model to obtain a second score and can identify the correlation between exercise performed by the user (1) and a disease.

[0079] FIG. 7 is an exemplary diagram illustrating a method for determining a correlation between medication information and disease possibility according to one embodiment of the present disclosure.

[0080] According to one embodiment of the disclosure, context information of a user (1) may include medication information of the user (1). In this case, the processor (110) may determine a relationship between the medication taken by the user (1) and the possibility of a disease, which is included in the medication information of the user (1).

[0081] Referring to FIG. 7, if the processor (110) determines that the user (1) has a high possibility of having a disease (or if the first score is identified as being equal to or greater than a first reference value), the processor (110) may obtain the user's (1) medication information. For example, the processor (110) may request a camera (specifically, a camera included in a computing device or a camera included in an external electronic device (e.g., a smart phone, a smart watch, etc.) worn by the user (1)) to take a picture of the medication taken by the user (1), thereby obtaining the medication image. Alternatively, the processor (110) may receive information on the medication taken by the user (1) from the user (1). Then, the processor (110) may analyze the obtained medication information to determine the type of medication taken by the user (1) and generate context data. Then, the processor (110) may input the context data and electrocardiogram data into a pre-trained second neural network model to obtain a second score and determine the correlation between the medication taken by the user (1) and the disease.

[0082] Meanwhile, the processor (110) can input the drug image into a pre-trained neural network model (the fourth neural network model) to identify the drug taken by the user (1). At this time, the fourth neural network model may be a model trained to extract the drug included in the drug image and identify the type of drug taken by the user (1). The fourth neural network model may be trained based on input data including a plurality of drug images and training data including label data in which segmentation of the drug included in the drug image and a label for the type of the drug are assigned. The fourth neural network model may also be implemented as a multi-layer perceptron (MLP), a convolutional neural network (CNN), a recurrent neural network (RNN), a generative adversarial network, etc.

[0083] In addition, according to one embodiment of the present disclosure, the context information of the user (1) may include the activity information of the user (1). At this time, the processor (110) may determine the sleep time and rest time of the user (1) included in the activity information of the user (1) to determine the correlation between the state of the user (1) and the possibility of a disease. Specifically, when the processor (110) determines that the state of the user (1) is highly likely to be affected by a disease (or when the first score is identified as being equal to or greater than a first reference value), the processor (110) may obtain the activity information of the user (1) through a sensor included in an external electronic device (e.g., a smart phone, a smart watch, etc.) worn by the user (1). The activity information may include information on the rest and sleep activities of the user (1), and the processor (110) may determine the correlation between the state of the user (1) and the possibility of a disease according to the rest time, sleep time, or work hours of the user (1). For example, if the processor (110) determines that the user (1) has a high probability of being ill, the processor (110) can generate context data by calculating the user's (1) recent resting or sleeping time. Then, the processor (110) can input the context data and electrocardiogram data into a pre-trained second neural network model to obtain a second score and determine the correlation between the medication taken by the user (1) and the disease. More specifically, if the probability of being ill is determined in the morning, the processor (110) can determine the correlation between the user's (1) condition and the probability of being ill based on the sleeping time, and if the probability of being ill is determined in the afternoon and evening, the processor (110) can determine the correlation between the user's (1) condition and the probability of being ill based on the rest time.

[0084] Referring back to FIG. 3, if the processor (110) determines that the acquired context information is related to a disease, it can provide guide information to the user (1). Specifically, if the second score is greater than or equal to a second reference value, the processor (110) can determine that the context information corresponding to the second score is related to a disease. On the other hand, if the second score is less than or equal to the second reference value, the processor (110) can determine that the context information corresponding to the second score is not related to a disease. If the processor (110) determines that the context information is related to a disease, it can provide the user (1) with guide information related to the context information. The guide information can be transmitted to an external electronic device (smartphone, smartwatch, etc.) used by the user (1) via a communication interface. For example, if the processor (110) determines that sodium levels affect left ventricular systolic dysfunction, it can provide the user (1) with guide information recommending diet adjustment or a low-salt diet. The guidance information can be transmitted to an external electronic device (smartphone, smartwatch, etc.) used by the user (1) via a communication interface. Alternatively, if the user's (1) high-intensity exercise is determined to affect left ventricular systolic dysfunction, guidance information may be provided to recommend that the user (1) adjust the exercise time or take a rest. Alternatively, if the user's (1) medication is determined to affect left ventricular systolic dysfunction, guidance information may be provided to recommend that the user (1) retake the medication prescription or adjust the number of medications taken. Alternatively, if the user's (1) condition due to insufficient sleep (or rest) is determined to affect left ventricular systolic dysfunction, guidance information may be provided to recommend that the user (1) sleep or adjust his / her lifestyle.

[0085] According to one embodiment of the present disclosure, when a plurality of context information of a user (1) is acquired, a plurality of second scores corresponding to each of the plurality of context information may be acquired, and based on the plurality of second scores, main context information having the highest correlation with a disease among the plurality of context information may be identified. Specifically, the processor (110) may periodically acquire context information. Alternatively, if the possibility of the user (1) being susceptible to a disease is determined to be high (or if the first score is identified as being equal to or higher than a first reference value), the processor may request acquisition of a plurality of different types of context information to acquire the plurality of context information. At this time, the processor (110) may calculate a second score corresponding to each of the plurality of context information, and identify the context information corresponding to the highest second score among the calculated second scores as main context information. At this time, the processor (110) may provide guide information for the user (1) related to the main context information. Alternatively, the guide information corresponding to the plurality of context information may be sequentially provided in the order of the second scores or provided in a list format.

[0086] Meanwhile, the processor (110) may determine the relationship with the possibility of a disease by using only the context information acquired within a preset time from the time of determining the possibility of a disease for the same type of context information.

[0087] FIG. 8 is an exemplary diagram illustrating the preparation of multiple first neural network models according to a user environment in which biometric data is acquired according to an embodiment of the present disclosure.

[0088] According to one embodiment of the present disclosure, the first neural network model and the first reference value, which is a criterion for determining the possibility of a disease, may be set in accordance with the environment in which the user's (1) biometric data was acquired. Here, the environment in which the biometric data was acquired may be the type of medical institution in which the biometric signal corresponding to the biometric data was measured or the type of biometric signal measuring device.

[0089] Specifically, the processor (110) can set the first neural network model and the first reference value according to the type of medical institution (or the type of measuring device) that acquired the biometric data of the user (1). In particular, the processor (110) can prepare a neural network model (hereinafter, a foundation model) learned based on a large-scale learning data set, and perform additional learning on the neural network model based on an additional learning data set corresponding to each device (and facility, institution, etc.) that acquired the biometric data.

[0090] At this time, additional learning can be performed through transfer learning. The processor (110) can utilize transfer learning techniques to optimize the foundation model to suit the data characteristics of a specific device (and facility, institution, etc.). Transfer learning is a machine learning technique that adapts the foundation model, previously trained on a large-scale dataset, to data collected in a new environment while maintaining the knowledge. Through this, the processor (110) can improve analysis accuracy by reflecting the characteristics of biometric data acquired from a specific device while maintaining the generalized knowledge of the existing foundation model. During the transfer learning process, additional learning can be performed using a new dataset while keeping some layers of the foundation model fixed. At this time, the layers responsible for low-level features of the foundation model are maintained, and only the layers responsible for high-level features are updated, thereby minimizing the loss of learned information in the existing model. Furthermore, if the characteristics of biometric data measured from a specific device type differ from the existing training data, the entire model can be fine-tuned for optimization. For example, electrocardiogram data acquired from medical institutions is relatively low in noise and highly reliable, whereas data acquired from wearable devices such as smartwatches may have poor signal quality or exhibit high variability due to environmental factors. Therefore, transfer learning can compensate for these differences and build a first-order neural network model appropriate for the specific environment.

[0091] Referring to Figure 8, multiple first neural network models can be prepared to correspond to tertiary hospitals, secondary hospitals, primary hospitals, nursing hospitals, health screening centers, and non-medical institutions (e.g., smartwatches). Here, a tertiary hospital may be a general hospital or university hospital, with specialists assigned to all medical departments and more than 500 beds. Tertiary hospitals may be determined based on whether they provide treatment for severe, rare, or incurable diseases, and whether they are equipped with state-of-the-art medical equipment and research facilities. Furthermore, tertiary hospitals provide treatment for patients referred from primary and secondary hospitals and can provide high-difficulty surgeries and detailed examinations. A secondary hospital may be a general hospital with at least four medical departments, two or more specialized departments, and 30 to 500 beds. Secondary hospitals handle diseases that are difficult to treat at primary hospitals and can provide emergency treatment and hospitalization. Primary hospitals are medical institutions at the level of clinics and may include community hospitals or private clinics. Primary hospitals primarily treat outpatients and may have fewer than 30 beds for short-term hospitalization. They treat general diseases and, when necessary, refer patients to secondary or tertiary hospitals. Nursing hospitals provide rehabilitation treatment and chronic disease management for patients requiring long-term care, and health screening centers are medical institutions that specialize in screening for the early detection and prevention of specific diseases. Non-medical institutions (e.g., smartwatches) may be devices that monitor vital signs and collect health data in locations other than medical institutions. The processor (110) may transfer-learn a foundation model based on electrocardiogram data acquired from each institution, thereby acquiring multiple first neural network models corresponding to each institution.At this time, the first reference value set for each first neural network model may also be set in accordance with the type of medical institution acquiring the biometric data. In particular, the higher the quality of the electrocardiogram data, the higher the first reference value may be set. Meanwhile, the type and number of medical institutions are not limited to those shown in FIG. 8 and may be set in various ways.

[0092] Accordingly, the processor (110) can prepare a plurality of first neural network models corresponding to the devices (and facilities, institutions, etc.) that acquired biometric data. Through this, the processor (110) can improve analysis accuracy by applying a first neural network model that reflects the characteristics of biometric data acquired from a specific device (and medical facility, medical institution, etc.). In particular, since each first neural network model is optimized by considering the distribution and characteristics of biometric data acquired from each device, deviations depending on the biometric data acquisition method are reduced, and more precise disease possibility assessment and health status analysis are possible.

[0093] Accordingly, when biometric data is acquired, the processor (110) can select a neural network model as the first neural network model, corresponding to the type of medical institution that acquired the user's biometric data, from among a plurality of neural network models trained to predict a disease based on a plurality of biometric data of the same type as the biometric data. At this time, the processor (110) can acquire location information of the user (1) acquired through a sensing unit (e.g., a GPS sensor), determine the location at the time when the user's biometric data was acquired based on the acquired location information, determine the type of medical institution that acquired the biometric data, and select the first neural network model (10).

[0094] Additionally, when a new device type is added, the processor (110) may utilize biometric data acquired from the device as an additional learning data set to update the foundation neural network model or generate a new first neural network model. In this case, the processor (110) may perform transfer learning using the existing learned foundation model, thereby efficiently acquiring the first neural network model while maintaining high accuracy even in a new environment.

[0095] In addition, when the processor (110) detects that the type of medical institution from which the user's biometric data is acquired has changed, the processor (110) selects a first neural network model corresponding to the new type of medical institution from among a plurality of neural network models, inputs the acquired user's biometric data into the pre-trained first neural network model, and obtains a first score corresponding to the possibility of a disease. That is, the processor (110) can dynamically determine an optimal neural network model capable of more accurately analyzing the biometric data by taking into account the type of medical institution from which the user's biometric data was acquired.

[0096] According to one embodiment of the present disclosure, if the first score is equal to or higher than the first reference value by a preset value, the acquired biometric data of the user (1) is assigned a first label value related to the possibility of a disease and a second label value corresponding to the type of the device, and set as learning data, and the learning data can be transmitted to another device of the same type as the type of the device. That is, if the processor (110) determines that the possibility of error in the judgment result of the possibility of a disease based on the biometric data is low, the processor (110) can assign a label to the biometric data and transmit it to another device of the same type as the device from which the biometric data was acquired so that it can be used as learning data. Of course, the processor (110) can also acquire biometric data from another device. At this time, the preset value can also be set differently depending on the type of medical institution.

[0097] FIG. 9 is a block diagram of a computing device (900) according to another embodiment of the present disclosure.

[0098] Referring to FIG. 9, a computing device (900) according to another embodiment of the present disclosure includes a processor (910), a memory (920), a communication interface (930), a sensing unit (940), a display (950), a user interface (960), a camera (970), and a speaker (980). Among the components illustrated in FIG. 9, the processor (910), the memory (920), and the communication interface (930) correspond to the processor (110), the memory (120), and the communication interface (130) of the computing device (100) illustrated in FIG. 2, and thus a detailed description thereof will be omitted.

[0099] A sensing unit (940) according to an embodiment of the present disclosure can obtain biometric information about a user (1). For example, the sensing unit (940) may include at least one electrode. At this time, the processor (910) can obtain an electrocardiogram signal of the user (1) through the at least one electrode. In addition, the sensing unit (940) may include an image sensor or an optical sensor. At this time, the processor (110) can obtain biometric information such as an optical blood flow signal through the image sensor (or optical sensor). In addition, the sensing unit (940) may further include a PPG sensor, a blood pressure sensor, a body temperature sensor, etc., through which various biometric information such as optical blood flow, blood pressure, body temperature, etc. about the user (1) can be obtained.

[0100] Meanwhile, the sensing unit (940) may include a GPS sensor, and at this time, the processor (910) may determine the type of medical institution that acquired the biometric data by determining the location at the time when the biometric data of the user (1) was acquired based on the location information of the user (1) acquired through the GPS sensor.

[0101] The display (950) can display various images. Here, the images include both still images and moving images. The display (950) can output guide information based on the diagnosis results of the user (1). The display (950) can be implemented as various types of displays, such as an LCD (Liquid Crystal Display Panel), an OLED (Organic Light Emitting Diodes), an LCoS (Liquid Crystal on Silicon), a DLP (Digital Light Processing), etc. In addition, the display (950) can also include a driving circuit, a backlight unit, etc., which can be implemented in a form such as an a-si TFT, an LTPS (low temperature poly silicon) TFT, an OTFT (organic TFT), etc. As an example, the display (950) can display the judgment results and guide information according to the first and second scores.

[0102] Meanwhile, the display (950) may be implemented as a touch screen by being combined with a touch panel, and in this case, the display (950) may perform the function of not only an output interface that outputs an image through the touch screen, but also an input interface that receives a touch input from a user (1).

[0103] The user interface (960) is a component used by the computing device (100) to perform interaction with the user (1), and may include at least one of a touch sensor, a motion sensor, a button, a jog dial, and a switch, but is not limited thereto. The processor (910) may receive input of diet information, exercise information, activity information, medication information, etc. through the user interface (960).

[0104] The camera (970) captures an image of an object around the user (1) and obtains an image of the object. Specifically, the camera (970) can capture an image of food consumed by the user (1). At this time, the processor (910) can determine the user's (1) diet information based on the image of the food consumed by the user (1), or can determine the user's (1) medication information based on the image of the medicine taken by the user (1). In addition, for this purpose, the camera (970) can be implemented with an imaging device such as an imaging device having a CMOS structure (CIS, CMOS Image Sensor) or an imaging device (Charge Coupled Device) having a CCD structure. However, the present invention is not limited thereto, and the camera (970) can be implemented with a camera module having various resolutions capable of capturing an object. Meanwhile, the camera (970) can be implemented with a depth camera (for example, an IR depth camera, etc.), a stereo camera, an RGB camera, etc.

[0105] The speaker (980) is a component that outputs various audio data on which various processing operations such as decoding, amplification, and noise filtering have been performed by an audio processing unit (not shown). The speaker (980) can output various notification sounds or voice messages. According to one embodiment of the present disclosure, the processor (910) can convert an electrical signal received from an external device into a user's (1) voice and output it through the speaker (980). For example, the speaker (980) can output user's (1) guide information related to context information as a voice message.

[0106] The microphone (990) can receive the user's voice. The processor (910) can analyze the user's voice or surrounding sounds obtained through the microphone (990) to obtain context information related to the user (1).

[0107] The various embodiments of the present disclosure described above can be combined with additional embodiments and modified within the scope understood by those skilled in the art in light of the detailed description above. It should be understood that the embodiments of the present disclosure are illustrative in all respects and not restrictive. For example, each component described as a single component may be implemented in a distributed manner, and likewise, components described as distributed may be implemented in a combined manner. Accordingly, all changes or modifications derived from the meaning, scope, and equivalent concepts of the claims of the present disclosure should be construed as being included within the scope of the present disclosure.

Claims

1. A method for building a neural network model suitable for a user environment, performed by a computing device including at least one processor, A step of inputting the acquired user's biometric data into a pre-trained first neural network model to obtain a first score corresponding to the possibility of a disease; and A step of predicting the user's status based on the first score; The above first neural network model is, Among the plurality of second neural network models trained to predict the condition based on a plurality of biometric data of the same type as the biometric data, one selected in response to the type of medical institution that acquired the user's biometric data. method.

2. In paragraph 1, The step of predicting the status of the above user is: If the first score is greater than or equal to a first reference value, a step of predicting that the user is included in a high-risk group for the disease is included; The above first reference value is, It is set in response to the type of place where the biometric data of the above user was acquired. method.

3. In paragraph 2, If the first score is higher than the first reference value by a preset value, a step of assigning a first label value related to the possibility of the disease to the acquired user's biometric data and setting it as learning data, and transmitting the learning data to another device that uses the same neural network model as the first neural network model; method.

4. In paragraph 1, A step of obtaining a foundation model by performing primary learning based on a first learning data set including a plurality of biometric data obtained from a plurality of users; and A step of constructing a second learning data set corresponding to each type of medical institution based on biometric data acquired from at least one medical institution corresponding to each type of medical institution, and performing secondary learning on the previously learned foundation model based on the second learning data set to obtain the plurality of second neural network models; comprising; method.

5. In paragraph 4, The above secondary learning is transfer learning performed based on the second learning data set. method.

6. In paragraph 1, If it is detected that the type of medical institution that acquires the user's biometric data has changed, a step of selecting a second neural network model corresponding to the new type of medical institution from among the plurality of second neural network models, and inputting the acquired user's biometric data into the pre-trained second neural network model to acquire a first score corresponding to the possibility of a disease; including; method.

7. In paragraph 6, A step of identifying the location of the user at the time when the biometric data is acquired through a sensor, and determining the type of medical institution that acquired the biometric data of the user based on the identified location; method.

8. In paragraph 1, The above plurality of second neural network models are, Including a neural network model corresponding to a tertiary hospital, a neural network model corresponding to a secondary hospital, a neural network model corresponding to a primary hospital, a neural network model corresponding to a nursing hospital, a neural network model corresponding to a health screening center, and a neural network model corresponding to a non-medical institution. method.

9. In paragraph 7, If the computing device is a smartwatch, the first neural network model is selected as a neural network model corresponding to a non-medical institution. method.

10. A computer program stored in a computer-readable storage medium, wherein the computer program, when executed on one or more processors, performs an operation of constructing a neural network model suitable for a user environment. The above action is, An operation of inputting the acquired user's biometric data into a pre-trained first neural network model to obtain a first score corresponding to the possibility of a disease; and An operation for predicting the user's status based on the first score; The above first neural network model is, Among the plurality of second neural network models trained to predict the condition based on a plurality of biometric data of the same type as the biometric data, one selected in response to the type of medical institution that acquired the user's biometric data. Computer program.

11. In a computing device that builds a neural network model suitable for the user environment, A communication interface for obtaining the user's biometric data; memory containing program codes; and It includes one or more processors that input the user's acquired biometric data into a pre-trained first neural network model to obtain a first score corresponding to the possibility of a disease, and predict the user's condition based on the first score. The above first neural network model is, Among the plurality of second neural network models trained to predict the condition based on a plurality of biometric data of the same type as the biometric data, one selected in response to the type of medical institution that acquired the user's biometric data. Computing device.

Images