Information processing system, program, and information processing device

Abstract

An information processing system (1) of an embodiment has a determination section (102) that generates a determination result indicating the presence or absence and the degree of a cardiac disorder of a subject by inputting pulsation information related to the heartbeat of the subject and attribute information of the subject to a learned model. The learned model is a learned model obtained by learning, with respect to a plurality of providers of training data, pulsation information related to the heartbeat of the providers, attribute information of the providers, and diagnosis information indicating the cardiac functional state of the providers.

Description

Technical Field

[0001] The embodiments of the present invention relate to information processing systems, programs, and information processing devices. Background Technology

[0002] Heart disease is currently the second leading cause of death in Japan, including known types such as ischemic heart disease, valvular heart disease, cardiomyopathy, arrhythmia, and heart failure. Heart disease can be life-threatening once it develops, therefore early detection and treatment are crucial. Current medical practice classifies heart disease into multiple stages, and treatment plans are discussed based on these stages.

[0003] For example, heart failure, a type of heart disease, is classified into four stages, A through D, based on its severity. Then, based on the severity of the stage, measures such as lifestyle modifications, medication, and palliative care are implemented.

[0004] Existing technical documents

[0005] Patent documents

[0006] Patent Document 1: Japanese Patent No. 6893002 Summary of the Invention

[0007] The problem that the invention aims to solve

[0008] However, although heart disease requires early detection and treatment, a visit to a medical institution is necessary to determine its presence and severity. Therefore, by the time patients realize they have symptoms and seek medical attention, the condition may have already progressed.

[0009] The purpose of this invention is to provide an information processing system, program, and device that can easily determine the risk of heart disease.

[0010] Technical means to solve the problem

[0011] To address the aforementioned issues and achieve the objectives, the present invention provides an information processing system comprising a determination unit that generates a determination result indicating the presence and severity of heart disease in the subject by inputting pulsation information related to the subject's heartbeat and attribute information of the subject into a learned model. The learned model is obtained by learning from multiple providers who provide training data, using pulsation information related to the heartbeat of each provider, attribute information of each provider, and diagnostic information indicating the cardiac function status of each provider.

[0012] Additionally, the present invention is a program in which a computer performs a determination process that generates a determination result indicating the presence and severity of heart disease in the subject by inputting pulsatile information related to the subject's heartbeat and attribute information of the subject into a learned model. The learned model is a learned model obtained by learning from a plurality of providers who provide training data, using pulsatile information related to the heartbeat of the providers, attribute information of the providers, and diagnostic information indicating the cardiac function status of the providers.

[0013] Furthermore, the present invention is an information processing device, which includes a determination unit that generates a determination result indicating the presence and severity of heart disease in the subject by inputting pulsation information related to the subject's heartbeat and attribute information of the subject into a learned model. The learned model is a learned model obtained by learning from multiple providers who provide training data, using pulsation information related to the heartbeat of the providers, attribute information of the providers, and diagnostic information indicating the cardiac function status of the providers.

[0014] Furthermore, the present invention is an information processing apparatus comprising a learning unit that, in order to generate a learned model that outputs a determination result indicating the presence and severity of heart disease in a subject by inputting pulsation information related to the subject's heartbeat and attribute information of the subject, performs machine learning for a plurality of providers providing training data, using pulsation information related to the heartbeat of the providers, attribute information of the providers, and diagnostic information indicating the cardiac function status of the providers.

[0015] Invention Effects

[0016] According to the present invention, an information processing system, program, and information processing device that can easily determine the risk of heart disease can be provided. Attached Figure Description

[0017] Figure 1 This is a diagram illustrating an example of the general structure of an information processing system.

[0018] Figure 2 This is a diagram illustrating an example of the hardware structure of an information processing device.

[0019] Figure 3 This is a flowchart illustrating the processing steps of information processing system 1.

[0020] Figure 4 It is a diagram used to illustrate the calculation and processing of characteristic quantities.

[0021] Figure 5 This is a diagram used to illustrate how sample entropy is calculated.

[0022] Figure 6 This is a diagram used to illustrate the completed learning model of Information Processing System 1.

[0023] Figure 7 This is a diagram illustrating an example of the general structure of information processing system 2.

[0024] Figure 8A This is a graph showing the verification results of the machine learning model in Example 1.

[0025] Figure 8B This is a graph showing the verification results of the machine learning model in Example 1.

[0026] Figure 9A This is a graph showing the verification results of the machine learning model in Example 1.

[0027] Figure 9B This is a graph showing the verification results of the machine learning model in Example 1.

[0028] Figure 10 This is a diagram used to illustrate the generation of input data in Example 2.

[0029] Figure 11A This is a graph showing the verification results of the machine learning model in Example 2.

[0030] Figure 11B This is a graph showing the verification results of the machine learning model in Example 2.

[0031] Figure 12A This is a graph showing the verification results of the machine learning model in Example 2.

[0032] Figure 12B This is a graph showing the verification results of the machine learning model in Example 2.

[0033] Figure 13A This is a graph showing the verification results of the machine learning model in Example 2.

[0034] Figure 13B This is a graph showing the verification results of the machine learning model in Example 2. Detailed Implementation

[0035] The information processing system, program, and information processing apparatus of the present invention will now be described with reference to the accompanying drawings. Furthermore, the following embodiments are not limited to the descriptions provided. Additionally, each embodiment can be combined with other embodiments or existing technologies to the extent that there is no contradiction in the processing content.

[0036] (Implementation Method)

[0037] First, use Figure 1 The general structure of the information processing system 1 of the present invention will be described. Figure 1 This is a diagram illustrating an example of the general structure of information processing system 1. For example... Figure 1 As shown, the information processing system 1 of the embodiment includes an information processing device 10, a wearable sensor 20, and an operation terminal 30.

[0038] The information processing device 10 is a server device that provides a service for determining the risk of heart disease. For example, the information processing device 10 includes an acquisition unit 101, a determination unit 102, and an output control unit 103. Furthermore, the functions of the information processing device 10 are not limited to the acquisition unit 101, the determination unit 102, and the output control unit 103. The acquisition unit 101, the determination unit 102, and the output control unit 103 will be described later.

[0039] The wearable sensor 20 is formed in the shape of a flexible plate and is a sensor device installed on the torso of a subject. For example, the wearable sensor 20 is installed by adhering or attaching to the skin of the torso. Furthermore, the subject is someone whose presence and severity of heart disease are being determined, including, for example, patients with heart disease, suspected patients with heart disease, and healthy individuals.

[0040] For example, wearable sensor 20 is a multi-sensor device incorporating an electrocardiograph (ECG) and a kinesiometer. The ECG measures temporal ECG information (ECG waveform data) based on electrical signals flowing through the subject's body. The kinesiometer (a 3-axis accelerometer) measures temporal acceleration information (acceleration data) based on the subject's body movements. Wearable sensor 20 records the various measured data and measurement times in a corresponding manner in its internal memory. Thus, wearable sensor 20 detects the subject's ECG and acceleration information. Furthermore, wearable sensor 20 is an example of a detection unit.

[0041] The operating terminal 30 is an information processing device operated by an operator, such as a personal computer, workstation, smartphone, or tablet computer. The operating terminal 30 is interconnected with both the information processing device 10 and the wearable sensor 20 via any communication method. For example, any network such as a LAN (Local Area Network) or WAN (Wide Area Network) can be used. Furthermore, the operator is the person operating the operating terminal 30, such as a doctor or someone working in a medical institution. However, the operator is not limited to this. For example, the operator can be the recipient or someone assisting the recipient.

[0042] Here, regarding Figure 1A representative example of the processing in the information processing system 1 shown will be described. The wearable sensor 20 is worn by the subject for a certain period of several days to several weeks, collecting electrocardiogram (ECG) information during that period. When the subject needs to determine the presence and severity of heart disease, they remove the wearable sensor 20 and hand it over to the operator. The operator operates the operating terminal 30, accesses the memory inside the wearable sensor 20, and reads the ECG information collected over the period. Then, the operator sends the read ECG information to the information processing device 10. The information processing device 10 determines the presence and severity of heart disease based on the ECG information and sends the determination result to the operating terminal 30. The operator reviews the determination result sent from the information processing device 10 and takes various actions based on the determination result.

[0043] In addition, Figure 1 The description provided is merely an example, and the invention is not limited thereto. For instance, in addition to an electrocardiograph or a motion analyzer, the wearable sensor 20 may also include any type of sensor such as a pulse wave meter for measuring pulse waves or a thermistor for measuring temperature (body temperature). Furthermore, the wearable sensor 20 may not necessarily include a motion analyzer. That is, the wearable sensor 20 only needs to include an electrocardiograph for detecting electrocardiographic information.

[0044] Alternatively, for example, the operating terminal 30 may not be permanently connected to the information processing device 10 and the wearable sensor 20. It is sufficient for the devices to be connected only when exchanging the various types of information. Furthermore, when exchanging information via a recording medium or the like, the devices may not need to be connected. Additionally, the wearable sensor 20 may send various types of information directly to the information processing device 10 without going through the operating terminal 30.

[0045] Furthermore, the wearable sensor 20 is preferably always worn by the user, but it can also be removed without affecting the processing of the information processing device 10. In addition, various measurement data recorded in the wearable sensor 20 can also be automatically transmitted to the information processing device 10 periodically (e.g., every few hours) via the network.

[0046] Next, use Figure 2 The hardware structure of the information processing device 10 in the information processing system 1 will be described. Figure 2 This is a diagram illustrating an example of the hardware structure of the information processing device 10. Furthermore, in Figure 2 The hardware structure of the information processing device 10 will be described in this section, but the hardware structure of the operation terminal 30 is basically the same, so the description is omitted.

[0047] like Figure 2As shown, the information processing device 10 includes: a CPU (Central Processing Unit) 11, a ROM (Read Only Memory) 12, a RAM (Random Access Memory) 13, an auxiliary storage device 14, an input device 15, a display device 16, and an external I / F (Interface) 17.

[0048] CPU 11 is a processor (processing circuit) that controls the operation of information processing device 10 by executing programs and realizes the various functions of information processing device 10. For example, the functions of acquisition unit 101, determination unit 102 and output control unit 103 of information processing device 10 are realized by CPU 11.

[0049] ROM 12 is a non-volatile memory that stores various data (information written during the manufacturing stage of the information processing device 10), including the program used to start the information processing device 10. RAM 13 is a volatile memory with the operating area of ​​the CPU 11. Auxiliary storage device 14 stores various data, such as the program executed by the CPU 11. Auxiliary storage device 14 may be composed of, for example, HDD (Hard Disc Drive) or SSD (Solid State Drive).

[0050] Input device 15 is a device used by a person operating information processing device 10 to perform various operations. Input device 15 may be, for example, a mouse, keyboard, touch panel, or hardware key.

[0051] Display device 16 displays various information. For example, display device 16 displays image data, model data, GUI (Graphical User Interface) for handling various operations from the operator, medical images, etc. Display device 16 may be composed of, for example, a liquid crystal display, an organic EL (Electroluminescence) display, or a cathode ray tube display. Furthermore, for example, input device 15 and display device 16 may be integrated into one unit using a touch panel.

[0052] External I / F17 is an interface used to connect (communicate) with any external device such as server device 20.

[0053] In addition, Figure 2 The description provided is merely an example, and the present invention is not limited thereto. For example, the hardware structure of the information processing device 10 can be arbitrarily applied using the structure of a known computer or workstation, etc.

[0054] use Figure 3 This describes the processing of information by the information processing device 10 in the information processing system 1. Figure 3 This is a flowchart illustrating the processing steps of information processing system 1. Figure 3 The processing steps shown may begin, for example, according to the operator's requirements. Furthermore, in Figure 3 The explanation should be appropriately referenced. Figure 1 , Figure 2 and Figures 4-6 Please provide an explanation.

[0055] like Figure 3 As shown, the acquisition unit 101 of the information processing device 10 receives the electrocardiogram information and attribute information of the subject (step S101).

[0056] Here, the electrocardiogram (ECG) information is continuous data, which is the electrical potential measured by the electrocardiograph included in the wearable sensor 20 and recorded continuously as numerical data. If the wearable sensor 20 is worn by the subject for seven days, the subject's ECG information is equivalent to approximately 168 hours of continuous data. However, the subject's ECG information may not necessarily be 168 hours. For example, the subject's ECG information preferably includes more than 48 hours of continuous data, and more preferably more than 144 hours of continuous data.

[0057] In addition, the attribute information includes at least one of the subject's age, gender, height, weight, and BMI (Body Mass Index). BMI can be calculated using well-known methods. Furthermore, the subject's age and BMI are preferably used as attribute information.

[0058] For example, an operator operates the operating terminal 30 to read time-series electrocardiogram (ECG) information from a wearable sensor 20 worn by the subject for seven days (approximately 168 hours). Additionally, the operator operates the operating terminal 30 to input the subject's attribute information. Thus, the operating terminal 30 acquires the subject's ECG and attribute information. Then, the operating terminal 30 sends the acquired ECG and attribute information to the information processing device 10. The acquisition unit 101 of the information processing device 10 receives the time-series ECG and attribute information sent from the operating terminal 30.

[0059] Next, the acquisition unit 101 of the information processing device 10 acquires the subject's pulsation information based on the subject's electrocardiogram information (step S102). For example, the acquisition unit 101 calculates pulsation information related to the subject's heartbeat based on the temporal electrocardiogram information detected from the subject.

[0060] Specifically, the acquisition unit 101 generates a plurality of segmented data arranged in time sequence by dividing the electrocardiogram (ECG) information at predetermined intervals in the temporal direction. Next, the acquisition unit 101 calculates a plurality of statistical values ​​for each of the plurality of segmented data. Then, the acquisition unit 101 calculates pulsation information for each of the plurality of statistical values. Furthermore, pulsation information is an example of a feature quantity representing the temporal characteristics of the ECG information.

[0061] Here, use Figure 4 The calculation and processing of characteristic quantities are explained. Figure 4 This is a diagram used to illustrate the calculation and processing of characteristic quantities. For example... Figure 4 As shown, for example, electrocardiogram (ECG) information is a continuous dataset containing n numerical data points from "numerical data 1" to "numerical data n" (where n is a natural number). The acquisition unit 101 generates a plurality of segmented data points by sequentially dividing the n numerical data points contained in this continuous dataset into segments at 5-minute intervals, starting from the first segment. For example, when there are 100 numerical data points within a 5-minute interval, the acquisition unit 101 designates "numerical data 1" to "numerical data 100" as "segmented data 1" and "numerical data 101" to "numerical data 200" as "segmented data 2". In this way, the acquisition unit 101 generates m segmented data points from "segmented data 1" to "segmented data m" (where m is a natural number less than n).

[0062] Next, the acquisition unit 101 calculates a complex number of statistical values ​​for each of the m segmented data. As a complex number of statistical values, any type of statistical value can be arbitrarily selected, such as mean, variance, standard deviation, maximum value, minimum value, skewness, etc. For example, the acquisition unit 101 calculates the average of 100 numerical data points contained in segmented data 1 as statistical value data 1, and calculates the average of 100 numerical data points contained in segmented data 2 as statistical value data 2. In this way, the acquisition unit 101 calculates m statistical value data points (average data points) from "statistical value data 1" to "statistical value data m" by calculating the average of each of the m segmented data points. Therefore, the acquisition unit 101 calculates m statistical value data points for one type of statistical value.

[0063] Then, the acquisition unit 101 calculates the median and interquartile range for each of the multiple statistical values. For example, the acquisition unit 101 uses m statistical values ​​corresponding to the average values ​​of m segmented data to calculate the median or interquartile range. As a result, the acquisition unit 101 calculates the median or interquartile range as pulsation information based on the electrocardiogram information as continuous data in a time sequence. Furthermore, since the median and interquartile range calculated here are calculated based on segmented data divided along the time sequence, it can be said that the temporal direction characteristics of the electrocardiogram information are shown.

[0064] Here, the statistical values ​​include at least one of various statistical values ​​of heart rate variability obtained from electrocardiogram information, the peak frequency of a specified frequency band component in the power spectrum based on frequency analysis of heart rate variability, the integral value of the specified frequency band component, and the value based on nonlinear analysis of heart rate variability. Additionally, the pulsatility information includes at least one of the median and interquartile range of the statistical values.

[0065] Specifically, pulsation information includes at least one of the following: information related to the entropy of heartbeat variations, the average heartbeat interval in heartbeat variations, heart rate based on electrocardiogram information, information related to the integral value or peak frequency of the HF (High Frequency) component based on the frequency analysis of heartbeat variations, information related to the integral value or peak frequency of the LF (Low Frequency) component based on the frequency analysis of heartbeat variations, and information related to the standard deviation of the Lorenz curve of heartbeat variations.

[0066] In this invention, hundreds of characteristic quantities (pulsation information) can be calculated, but the useful characteristic quantities are shown in Table 1. In Table 1, "characteristic quantity" shows the types of useful pulsation information used to determine the presence and severity of heart disease. In addition, "remarks" show the meaning of each characteristic quantity.

[0067] Table 1

[0068] For example, the interquartile range of the sample entropy of heart rate variability is an indicator of the disorder (unpredictability) of temporal variations in the R-wave interval (RR interval: RRI). Sample entropy is a value calculated based on the number of sample groups and represents the degree of disorder in continuous values ​​where the variation pattern of the values ​​in that sample group is the same as that of a group (sample set) comprising any number of consecutive samples. Furthermore, sample entropy is an example of a statistical value.

[0069] For example, sample entropy is calculated using the following mathematical formula 1. In formula 1, SampEn represents sample entropy. m represents the number of samples in a sampling group. r represents the width (range) of values ​​used to determine whether the values ​​of the samples are close. n represents the number of samples. Ai represents the number of sampling groups in which the values ​​of three consecutive samples in a sampling group have the same change pattern. Bi represents the number of sampling groups in which the values ​​of two consecutive samples in a sampling group have the same change pattern.

[0070]

Mathematical Formula 1

[0071] Here, use Figure 5 The method for calculating sample entropy is explained. Figure 5 This is a diagram illustrating how sample entropy is calculated. Figure 5 The example demonstrates the derivation process of Ai and Bi in mathematical formula 1 using arbitrary continuous values ​​contained in the RRI timing sequence. Figure 5 In the diagram, the vertical axis corresponds to the magnitude of the RRI value, and the horizontal axis corresponds to the elapsed time. Furthermore, for... Figure 4 The sample entropy is calculated for each segment of data.

[0072] exist Figure 5 In the upper part, the acquisition unit 101 derives Bi of mathematical formula 1 by counting the number of sampling groups that have the same value change pattern as the sampling group SGB0, which includes two consecutive samples i and i+1. Figure 5 In the upper part, range R1 is the range of values ​​that can be determined to be close to the value of sample i. Additionally, range R2 is the range of values ​​that can be determined to be close to the value of sample i+1. The width (size) of both ranges R1 and R2 is r.

[0073] For example, the acquisition unit 101 determines sample j as a sample included in range R1. Then, the acquisition unit 101 determines whether the value of the next sample j+1 of the determined sample j is included in range R2. In this case, since the value of sample j+1 is included in range R2, the acquisition unit 101 determines that the change pattern of the value of the sampling group SGB1, which includes two consecutive samples j and j+1, is the same as that of the sampling group SGB0.

[0074] Furthermore, the acquisition unit 101 determines sample k as a sample included in range R1. Then, the acquisition unit 101 determines whether the value of the next sample k+1 after the determined sample k is included in range R2. In this case, since the value of sample k+1 is included in range R2, the acquisition unit 101 determines that the variation pattern of the value of the sampling group SGB1, which includes two consecutive samples k and k+1, is the same as that of the sampling group SGB0.

[0075] Then, the acquisition unit 101 performs the same processing to determine four sampling groups SGB1 to SGB4, which are sampling groups with the same value change pattern as sampling group SGB0. The number of sampling groups determined is "4", so the acquisition unit 101 derives "4" as Bi.

[0076] exist Figure 5 In the lower part, a sample is added as the object of analysis for the derivation processing of Bi. By comparing it with the variation pattern of the sampling group SGA0, which also includes sample i+2, Ai of mathematical formula 1 is derived. Figure 5In the lower part, range R1 is the range of values ​​that can be determined to be close to the value of sample i. Additionally, range R2 is the range of values ​​that can be determined to be close to the value of sample i+1. Furthermore, range R3 is the range of values ​​that can be determined to be close to the value of sample i+2. The width (size) of the values ​​in ranges R1, R2, and R3 is all r.

[0077] For example, the acquisition unit 101 determines whether the value of sample j+2 is included in the range R3. In this case, since the value of sample j+2 is not included in the range R3, the acquisition unit 101 determines that the change pattern of the sample group including three consecutive samples j, j+1, and j+2 is different from that of sample group SGA0.

[0078] Furthermore, the acquisition unit 101 determines whether the value of sample k+2 is included within the range R3. In this case, since the value of sample k+2 is included within the range R3, the acquisition unit 101 determines that the variation pattern of the value of the sampling group SGA1, which includes three consecutive samples k, k+1, and k+2, is the same as that of the sampling group SGA0.

[0079] Then, the acquisition unit 101 performs the same processing to determine two sampling groups SGA1 to SGA2, which are sampling groups whose value change patterns are the same as those of sampling group SGA0. The number of sampling groups determined is "2", so the acquisition unit 101 derives "2" as Ai.

[0080] Thus, the acquisition unit 101 derives Ai and Bi. Then, the acquisition unit 101 calculates the sample entropy of heart rate fluctuations as a statistical value by inputting the derived Ai and Bi into mathematical formula 1. The acquisition unit 101 calculates the interquartile range of the calculated sample entropy of heart rate fluctuations as pulsation information.

[0081] In addition, Figure 5 The description provided is merely an example, and the invention is not limited thereto. For instance, the number of samples included in a sampling group is not limited to [specific number]. Figure 5 The values ​​shown can be set to any value.

[0082] Return to the explanation in Table 1. For example, the interquartile range of the standard deviation of heart rate obtained from electrocardiogram information is an indicator of the change in heart rate deviation.

[0083] Additionally, the median SD2 / SD1 ratio of heart rate variability is an indicator of sympathetic nervous system activity. SD1 represents the standard deviation of the Lorenz curve of the RRI along the -x axis. SD2 represents the standard deviation of the Lorenz curve of the RRI along the -x axis.

[0084] In addition, the interquartile range of SD2 / SD1 for heart rate variability is an indicator of the deviation of the sympathetic nervous system.

[0085] In addition, the median CVI (Cardiac Vagal Index) of heart rate variability is an indicator of the parasympathetic nervous system. CVI is equivalent to the logarithm of the product of SD1 and SD2 (logarithm is commonly used).

[0086] Furthermore, the interquartile range of the LF (Low Frequency) correction value for heart rate variability is an indicator of the deviation from the sympathetic nervous system's parameters. The LF correction value is obtained by dividing the sum of LF and HF (High Frequency) by LF. Moreover, LF and HF are calculated based on the frequency analysis of heart rate variability. For example, LF is obtained by integrating the power of the LF component (0.04Hz–0.15Hz) in the power spectrum calculated from the RRI time series. Similarly, HF is obtained by integrating the power of the HF component (0.15Hz–0.40Hz) in the power spectrum calculated from the RRI time series.

[0087] In addition, the median peak frequency of the LF component of heart rate fluctuations is an indicator of the blood pressure fluctuation cycle.

[0088] In addition, the interquartile range of the mean heart rate variation is an indicator of the deviation of the time-series variation of the RRI.

[0089] In addition, the interquartile range of SD2 for heart rate variability is an indicator of the deviation from the total power of autonomic nervous system activity.

[0090] In addition, the interquartile range of the peak frequency of the HF component of heart rate variability is an indicator of the deviation from the respiratory cycle.

[0091] In addition, the median integral value of the LF component of heart rate variability is an indicator that includes both parasympathetic and sympathetic components.

[0092] Furthermore, the interquartile range of the approximate entropy of heart rate variability is an indicator of the disorder of RRI time-series variations. Moreover, the approximate entropy is calculated differently from the sample entropy and can be calculated using any of the well-known methods.

[0093] In addition, the median logarithm of the HF component of heart rate variability is an indicator after logarithmic transformation of the parasympathetic nervous system index. Furthermore, the logarithm of the HF component can be obtained, for example, using a common logarithm.

[0094] In addition, the median LF correction value for heart rate variability is an indicator of the sympathetic nervous system. Furthermore, the calculation method for the LF correction value is the same as described above, therefore, its explanation is omitted.

[0095] Furthermore, the content described in Table 1 is merely an example, and the present invention is not limited thereto. For example, the acquisition unit 101 may acquire at least one of the feature quantities shown in Table 1. However, as feature quantities, it is preferable to use the interquartile range of the sample entropy of heart rate variation, the interquartile range of the standard deviation of heart rate, the median of SD2 / SD1 of heart rate variation, the interquartile range of SD2 / SD1 of heart rate variation, the median of CVI (Cardiac Vagal Index) of heart rate variation, the interquartile range of the LF (Low Frequency) correction value of heart rate variation, and the median of the peak frequency of the LF component of heart rate variation.

[0096] In addition to the features shown in Table 1, any feature can be specified using any calculation method. Furthermore, frequency analysis includes at least one of high-speed Fourier transform, Lomb-Scargle periodogram, and autoregressive model. However, the pulsation information differs from the electrocardiogram waveform contained in electrocardiogram information.

[0097] return Figure 3 The determination unit 102 of the information processing device 10 generates a determination result by inputting pulsation information and attribute information into the learned model (step S103). Furthermore, this learned model is obtained by learning from a plurality of providers who provide training data, using pulsation information related to the provider's heartbeat, the provider's attribute information, and diagnostic information indicating the provider's cardiac function status. Furthermore, the providers include healthy individuals and patients diagnosed with heart failure by at least a doctor. That is, the determination result indicates the presence and degree of heart failure in the subject. The diagnostic information is information indicating the cardiac function status related to the provider's heart failure.

[0098] Here, use Figure 6 The learning model of Information Processing System 1 is explained. Figure 6 This is a diagram used to illustrate the completed learning model of Information Processing System 1. Figure 6 The upper part shows the processing during the learning process after the model has finished learning. Figure 6 The lower part shows the processing when the learned model is used. Here, the learned model is constructed in advance and stored in a predetermined storage area (e.g., ROM 12) that can be used by the determination unit 102. Furthermore, the processing of the "learning unit" used to construct the learned model will be explained later.

[0099] like Figure 6As shown in the upper part, during learning, machine learning is performed using, for example, the pulsation information, attribute information, and diagnostic information of a plurality of providers 1 to N. Here, the provider's pulsation information is calculated based on the provider's electrocardiogram information. The method for calculating the provider's pulsation information is the same as the method for calculating the pulsation information of the subject described in the acquisition unit 101, so the description is omitted. In addition, the provider's attribute information includes at least one of the provider's age, gender, height, weight, and BMI (Body Mass Index). BMI can be calculated using known calculation methods. In addition, the provider's diagnostic information is information indicating the provider's cardiac function status, which is the result of a doctor's diagnosis.

[0100] For example, a fully learned model can be constructed by performing XGBoost-based machine learning. This fully learned model takes the subject's pulsation information and attribute information as input, and outputs a judgment result indicating the presence and severity of the subject's heart disease.

[0101] Then, as Figure 6 As shown in the lower part, in use, the determination unit 102 inputs the subject's pulsation information and attribute information into the learned model constructed by machine learning, thereby causing the learned model to output a determination result. Thus, the determination unit 102 generates a determination result.

[0102] In addition, Figure 6 The learning-time processing described herein can be executed at any time interval, as long as it precedes the application-time processing. Furthermore, the learning-time processing can also be executed to update (append-learn) the already generated, learned model.

[0103] Furthermore, in this invention, the information input into the learned model is merely pulsation information or attribute information, not the values ​​of the electrocardiogram (ECG) information itself. That is, the pulsation information is different from the ECG waveform contained in the ECG information.

[0104] Then, the output control unit 103 outputs the determination result (step S104). For example, the output control unit 103 sends the determination result to the operation terminal 30. The operation terminal 30 displays the determination result sent from the output control unit 103 on a display device or stores it in a designated storage device.

[0105] As described above, in the information processing system 1 of the present invention, the determination unit 102 of the information processing device 10 generates a determination result indicating the presence and severity of heart disease in the subject by inputting pulsation information related to the subject's heartbeat and the subject's attribute information into the learned model. Thus, the information processing system 1 can easily determine the risk of heart disease.

[0106] For example, by utilizing information processing system 1, users can obtain risk assessment results for heart disease based on a learned model even without visiting a medical institution. As a result, information processing system 1 is expected to contribute to the early detection and early treatment of heart disease.

[0107] (Modified example)

[0108] In the above embodiments, the case where continuous data over a certain period of time obtained from the subject is used as the processing object for feature quantity calculation processing is described, but the present invention is not limited to this. For example, the acquisition unit 101 can select numerical data corresponding to sleep periods from the acquired continuous data over a certain period of time for feature quantity calculation processing.

[0109] That is, the acquisition unit 101 determines the sleep period based on at least one of a preset time, time-series electrocardiogram information detected from the subject, time-series acceleration information detected from the subject, and time-series posture information detected from the subject. Then, the acquisition unit 101 calculates pulsation information based on the electrocardiogram information of the sleep period in the time-series electrocardiogram information.

[0110] For example, the time used to determine sleep periods is sometimes preset from 0:00 AM to 6:00 AM. In this case, the acquisition unit 101 determines the time from 0:00 AM to 6:00 AM each day as the sleep period. Then, the acquisition unit 101 extracts the electrocardiogram information contained in the time from 0:00 AM to 6:00 AM each day from the time-series electrocardiogram information, and calculates pulsation information based on the extracted electrocardiogram information.

[0111] Additionally, for example, the acquisition unit 101 determines a sleep period based on temporal acceleration information detected from the subject. Here, the acceleration information is collected by a 3-axis accelerometer included in the wearable sensor 20. As an example, the acquisition unit 101 infers the subject's movements (body movements) based on the acceleration information. Then, if the magnitude of the inferred movement is less than a threshold for more than 3 hours, the acquisition unit 101 determines that time as a sleep period. Then, the acquisition unit 101 extracts the electrocardiogram (ECG) information contained in the determined time from the temporal ECG information and calculates pulsation information based on the extracted ECG information.

[0112] Additionally, for example, the acquisition unit 101 determines the sleep period based on temporal posture information detected from the subject. Here, the posture information is collected by a gyroscope sensor provided with the wearable sensor 20. As an example, the acquisition unit 101 infers the subject's posture based on the posture information. Then, the acquisition unit 101 determines the time the subject lies down as the sleep period. Then, the acquisition unit 101 extracts the electrocardiogram (ECG) information contained in the determined time from the temporal ECG information and calculates pulsation information based on the extracted ECG information.

[0113] In this way, the information processing device 10 can select electrocardiogram information corresponding to the sleep period from continuous data over a certain period and use it for characteristic quantity calculation. As a result, the information processing device 10 can make a judgment based on the collected conditions related to the physical movement of the subject or provider, thus improving the accuracy of the judgment.

[0114] (Other implementation methods)

[0115] In addition to the embodiments described above, the present invention can also be implemented in various other forms.

[0116] (System Structure)

[0117] In the above embodiments, using Figure 1 The general structure of the information processing system 1 has been described, but the present invention is not limited thereto. For example, the processing functions of the acquisition unit 101, the determination unit 102, and the output control unit 103 of the information processing device 10 can also be provided in any device of the information processing system 1. As an example, the acquisition unit 101 can also be provided in the operation terminal 30.

[0118] That is, the information processing system 1 has a determination unit that generates a judgment result determining the risk of heart disease by inputting pulsation information related to the heartbeat of the subject and the subject's attribute information into the learned model. The learned model is a learned model obtained by learning from multiple providers who provide training data, using pulsation information related to the heartbeat of the providers, the attribute information of the providers, and diagnostic information representing the cardiac function status of the providers.

[0119] (program)

[0120] Furthermore, each of the processes of the information processing system 1 described in the above embodiments and variations can also be provided as a program that causes a computer to execute these processes.

[0121] In other words, the program instructs the computer to perform a decision-making process. This process involves inputting pulsatile information related to the subject's heartbeat and the subject's attribute information into a fully learned model, thereby generating a determination of the presence and severity of the subject's heart disease. The fully learned model is obtained by learning from multiple providers who provide training data, using pulsatile information related to the providers' heartbeats, the providers' attribute information, and diagnostic information representing the providers' cardiac function status.

[0122] (Information processing device)

[0123] Furthermore, in the information processing system 1 described in the above embodiments and variations, the information processing device 10 can easily determine the risk of heart disease.

[0124] That is, the information processing device 10 has a determination unit that generates a determination result indicating the presence and severity of heart disease in the subject by inputting pulsation information related to the subject's heartbeat and the subject's attribute information into the learned model. The learned model is a learned model obtained by learning from multiple providers who provide training data, using pulsation information related to the provider's heartbeat, the provider's attribute information, and diagnostic information indicating the provider's cardiac function status.

[0125] Furthermore, the processing function of the wearable sensor 20, which serves as the detection unit, can also be equipped in the information processing device 10.

[0126] (Study Department)

[0127] In addition, the system of the present invention may also have a processing function for constructing the learned model described in the above embodiments.

[0128] use Figure 7 The general structure of the information processing system 2 of the present invention will be described. Figure 7 This is a diagram illustrating an example of the general structure of information processing system 2. For example... Figure 7 As shown, the information processing system 2 of the embodiment includes a wearable sensor 20, an operation terminal 30, and an information processing device 40. Figure 7 The structure of the wearable sensor 20 and the operating terminal 30 shown is similar to Figure 1 The wearable sensor 20 and the operating terminal 30 shown have basically the same structure, so the description is omitted.

[0129] The information processing device 40 is a server device that constructs a learned model for determining the risk of heart disease. For example, the information processing device 40 includes an acquisition unit 401, a learning unit 402, and an output control unit 403. Figure 7 The structure of the acquisition unit 401 and the output control unit 403 shown is similar to Figure 1 The acquisition unit 101 and the output control unit 103 shown have basically the same structure, so the description is omitted.

[0130] Here, the learning unit 402 uses pulsation information, attribute information, and diagnostic information from multiple providers 1 to N to perform machine learning, constructing a fully learned model for determining the presence and severity of heart disease in a subject. For example, as Figure 6As shown in the upper part, the learning unit 402 performs XGBoost-based machine learning. Through this machine learning, the learning unit 402 generates a learned model, which, given the subject's pulsation information and attribute information as input, outputs a judgment result indicating the presence and severity of the subject's heart disease. The learning unit 402 stores the generated learned model in any storage area that the judgment unit 102 can utilize.

[0131] That is, in the learning method of the present invention, in order to generate a learned model that outputs a judgment result indicating the presence and degree of heart disease in a subject by inputting pulsation information related to the subject's heartbeat and attribute information of the subject, machine learning is performed on multiple providers who provide training data using pulsation information related to the provider's heartbeat, attribute information of the provider, and diagnostic information indicating the provider's cardiac function status.

[0132] Furthermore, the case of performing XGBoost-based machine learning has been described here, but the present invention is not limited thereto. The learning unit 402 is not limited to XGBoost and can apply any known machine learning algorithm.

[0133] Furthermore, this description focuses on an information processing device 40, which differs from the information processing device 10, and includes a learning unit 402; however, the present invention is not limited to this. For example, the learning unit 402 may also be provided in the information processing device 10.

[0134] Based on the implementation methods and variations described above, the risk of heart disease can be easily determined.

[0135] Furthermore, the programs executed in the information processing systems 1 and 2 of the above embodiments and variations can also be provided as files recorded in an installable or executable form on a computer-readable recording medium such as a CD-ROM, floppy disk (FD), CD-R, DVD, or USB (Universal Serial Bus), or they can be provided or distributed via a network such as the Internet. Alternatively, they can be provided by pre-assembling various programs in a non-volatile storage medium such as a ROM.

[0136] [Example]

[0137] The present invention will now be described in more detail with reference to embodiments, but the present invention is not limited thereto.

[0138] (Example 1)

[0139] According to the method described in the above embodiments, two machine learning models (learned models) are created to classify heart failure into four stages, A to D, and their accuracy is verified.

[0140] Here, the classification of heart failure stages is explained. For example, according to the ACCF / AHA (American College of Cardiology Foundation / American Heart Association) stage classification, heart failure is divided into four stages, A through D. This classification is primarily based on the presence or absence of "organic heart disease" such as cardiac hypertrophy or decreased cardiac output, and the presence or absence of "heart failure syndrome" such as chest pain, shortness of breath, or leg edema. Stage A is a stage with risk factors such as hypertension and diabetes, but no organic heart disease or heart failure syndrome. Stage B is a stage with organic heart disease but no heart failure syndrome. Stage C is a stage with both organic heart disease and heart failure syndrome. Stage D is a stage with both organic heart disease and heart failure syndrome, and is in a refractory (treatment-resistant, end-stage) stage. Furthermore, the stages of heart failure that can be classified in this application are not limited to the examples above; the definition of the stage classification can be arbitrarily set.

[0141] Table 2 shows the samples used in the two machine learning models (Machine Learning Model A and Machine Learning Model B) and their validation results. Both machine learning models were created using the Nystroem Kernel SVM Classifier.

[0142] Table 2

[0143] In Table 2, the patient group represents the number of data samples obtained from patients diagnosed with heart failure by doctors, and the healthy group represents the number of data samples obtained from healthy individuals. That is, the samples from stage A are included in the healthy group, and the samples from stages B to D are included in the patient group. Furthermore, the samples used in machine learning model A and the samples used in machine learning model B contain overlapping content.

[0144] The results in Table 2 show that both Recall and Specificity values ​​are high for either Machine Learning Model A or Machine Learning Model B, confirming that a machine learning model with sufficient performance has been created.

[0145] Next, the data from the samples used in machine learning are input into each machine learning model, and the validation results of the judgment results are output again, showing that... Figure 8A , Figure 8B , Figure 9A ,and Figure 9B middle. Figure 8A , Figure 8B , Figure 9A and Figure 9B This is a graph showing the verification results of the machine learning model in Example 1.

[0146] Figure 8A The graph shows the re-output score based on machine learning model A for each patient / healthy group. Figure 8B The graph shows the re-output scores based on machine learning model B for each patient / healthy group. Figure 8A and Figure 8B In the diagram, the vertical axis represents the re-output score from each machine learning model. This score is represented by a value from 0 to 1, with a value closer to "1" indicating a higher probability of being a patient and a value closer to "0" indicating a higher probability of being a healthy person. That is, the prediction threshold is 0.5. The horizontal axis represents the attribute of being a patient or a healthy person, with the value in parentheses corresponding to the sample size. Figure 8A and Figure 8B The data of each sample shown in Table 2 are input into the machine learning model A again, and the scores obtained as the judgment result (the scores are output again) are depicted separately according to attributes (patient group and healthy group) and illustrated as box plots.

[0147] Although Figure 8A and Figure 8B The results showed several samples that exceeded the prediction threshold, but good results were obtained in any machine learning model.

[0148] Figure 9A The graph shows the re-output score based on machine learning model A for each stage. Figure 9B The graph shows the re-output score based on machine learning model B for each stage. Figure 9A and Figure 9B In the diagram, the vertical axis represents the re-output scores from each machine learning model. The scores are explained below. Figure 8A and Figure 8B Similarly, the horizontal axis represents the stage of heart failure diagnosed by doctors, and the values ​​in parentheses correspond to the sample size. Furthermore, Figure 9B The "unclear" value on the horizontal axis represents samples that, although diagnosed with heart failure, have not been classified into different stages of heart failure. That is, Figure 9A and Figure 9B The data from each sample shown in Table 2 are input into the machine learning model B again, and the scores obtained as the judgment result (the scores are output again) are plotted separately according to each stage of heart failure and illustrated as box plots.

[0149] Although Figure 9A and Figure 9BThe results showed that several samples exceeded the prediction threshold in stages A and C, but good results were obtained in any machine learning model. In particular, no samples exceeded the prediction threshold in stage B, and the model was able to classify the samples with high accuracy.

[0150] Heart failure is generally a difficult disease to treat, so early detection and treatment are crucial. However, by the time patients become aware of their condition due to heart failure syndrome, it has usually progressed to stage C, making early detection in stage B difficult. Figure 9A and Figure 9B The results show that stage B can be classified with high accuracy, demonstrating the effectiveness of the information processing device 10 shown in this embodiment. Therefore, by wearing the wearable sensor 20 to measure data for purposes such as health diagnosis, and determining the stage of heart failure using the information processing device 10 of this embodiment, it is possible to detect the condition at an early stage, before heart failure syndrome has developed.

[0151] (Example 2)

[0152] Next, the number of measurement days (measurement days) based on wearable sensor 20 was verified. In Example 2, a machine learning model was created using the same samples as those used in machine learning model A in Example 1, through the method described in the above embodiments.

[0153] Figure 10 This is a diagram used to illustrate the generation of input data in Example 2. For example... Figure 10 As shown, with a total of seven days of measurement data, seven input data sets were generated: one day's data, two days' data, three days' data, four days' data, five days' data, six days' data, and seven days' data. Each input data set is the number of days counted from the start of the measurement, extracting continuous data from the sleep period (i.e., from 0:00 to 6:00). Therefore, for example, the input data for three days' data includes: continuous data from 0:00 to 6:00 on the first day; continuous data from 0:00 to 6:00 on the second day; and continuous data from 0:00 to 6:00 on the third day.

[0154] Furthermore, depending on the sample (subject), sometimes it is impossible to obtain seven days' worth of measurement data due to reasons such as the wearable sensor 20 falling off within seven days. Therefore, for example, if a total of three days' worth of measurement data is obtained, three input data points are generated: one day's, two days', and three days' worth. Alternatively, for example, if a total of five days' worth of measurement data is obtained, five input data points are generated: one day's, two days', three days', four days', and five days' worth. Moreover, the samples used for the input data are the same samples used in machine learning model A in Example 1. The generated input data for each day's worth of data is input into the created machine learning model, and scores are output according to the segmented number of days.

[0155] Figure 11A , Figure 11B , Figure 12A , Figure 12B , Figure 13A and Figure 13B This is a graph showing the verification results of the machine learning model in Example 2.

[0156] Figure 11A The curve represents the score obtained from the sample of the patient population, expressed as a percentage of the number of days measured. Figure 11B The graph represents the score obtained from the healthy population sample, expressed as a percentage of the number of days measured. Figure 11A and Figure 11B In the graph, the vertical axis represents the score output from the machine learning model. The score is explained in conjunction with... Figure 8A and Figure 8B Same. The horizontal axis corresponds to the number of days measured. That is, Figure 11A and Figure 11B It is a line graph that plots the scores obtained as the judgment result according to the number of measurement days and connects the plotted points from the same sample with lines.

[0157] Figure 11A and Figure 11B The results show that, in either the patient group or the healthy group, the score fluctuates more as the number of measurement days decreases, and deviations from the prediction threshold (0.5) are also frequently observed.

[0158] Figure 12A The graph shows Figure 11A The fluctuations in scores for each measurement day are shown. Figure 12B The graph shows Figure 11B The fluctuations in scores for each measurement day are shown. Figure 13A The diagram is Figure 12A The graph shown is the result of converting the curve graph into a box plot. Figure 13B The diagram is Figure 12B The graph shown is the result of converting the curve to a box plot. Figures 12A to 13B In the diagram, the vertical axis corresponds to the score for the maximum number of measurement days for each subject (patient and healthy) and the difference (fluctuation) between the scores for each measurement day. That is, this difference assumes the score for the maximum number of measurement days for each subject is correct. The horizontal axis corresponds to the number of measurement days.

[0159] Figures 12A to 13B The results showed that, in either the patient or healthy population, the fluctuations were significant in the one-day and two-day doses, but the fluctuations were relatively smaller in the three-day and four-day doses, and even smaller in doses of five days or more. Based on these results, it is preferable to perform measurements for three days or more, and more preferably for five days or more.

[0160] In addition, the present invention provides technical means for solving the problem as described in the following appendix.

[0161] (Note 1) An information processing system, wherein, It has a determination unit that, by inputting pulsation information related to the subject's heartbeat and the subject's attribute information into the learned model, generates a determination result indicating the presence and severity of the subject's heart disease. The learned model is a learned model obtained by learning from multiple providers who provide training data, using pulsation information related to the heartbeat of the provider, attribute information of the provider, and diagnostic information representing the cardiac function status of the provider.

[0162] (Appendix 2) According to the information processing system described in Appendix 1, wherein, The providers include healthy individuals and patients diagnosed with heart failure by a doctor.

[0163] (Appendix 3) According to the information processing system described in Appendix 1, wherein, It also has a unit for acquiring the pulsation information of the subject.

[0164] (Appendix 4) According to the information processing system described in Appendix 1, wherein, The attribute information of the subject includes at least one of the subject's age, gender, height, weight, and BMI (Body Mass Index).

[0165] (Appendix 5) According to the information processing system described in Appendix 3, wherein, The acquisition unit calculates the pulsation information based on the temporal electrocardiogram information detected from the subject.

[0166] (Note 6) According to the information processing system described in Note 5, wherein, The acquisition unit, By segmenting the electrocardiogram (ECG) information at predetermined intervals along the temporal direction, a plurality of segmented data are generated in a temporal sequence. For each of the multiple segmented data, calculate a multiple number of statistical values. For each of the multiple statistical values, the pulsation information representing the temporal direction characteristics of the electrocardiogram information is calculated.

[0167] (Note 7) According to the information processing system described in Note 5, wherein, The electrocardiogram information includes more than 72 hours of continuous data obtained from the subject.

[0168] (Note 8) According to the information processing system described in Note 5, wherein, The electrocardiogram information includes more than 120 hours of continuous data obtained from the subject.

[0169] (Note 9) According to the information processing system described in Note 6, wherein, The statistical values ​​include at least one of the following: various statistical values ​​of heart rate variability obtained from the electrocardiogram information, the peak frequency of a specified frequency band component in the power spectrum based on the frequency analysis of the heart rate variability, the integral value of the specified frequency band component, and the value obtained from the nonlinear analysis of the heart rate variability. The pulsation information includes at least one of the median and interquartile range of the statistical values.

[0170] (Note 10) According to the information processing system described in Note 9, wherein, The attribute information of the subject is the subject's age and BMI (Body Mass Index). The pulsation information is the interquartile range of the sample entropy of the heartbeat variation.

[0171] (Note 11) According to the information processing system described in Note 10, wherein, The pulsation information also includes the median of the LF (Low Frequency) correction value based on the frequency analysis of the heart rate variation, the interquartile range of the standard deviation of the heart rate obtained from the electrocardiogram information, the median of the peak frequency of the LF component based on the frequency analysis of the heart rate variation, the median of the logarithm of the HF (High Frequency) component based on the frequency analysis of the heart rate variation, and the interquartile range of the LF correction value based on the frequency analysis of the heart rate variation.

[0172] (Note 12) According to the information processing system described in Note 9, wherein, The pulsation information includes at least one of the following: information related to the entropy of the heartbeat variation, the average heartbeat interval in the heartbeat variation, the heart rate based on the electrocardiogram information, information related to the integral value or peak frequency of the HF (High Frequency) component based on the frequency analysis of the heartbeat variation, information related to the integral value or peak frequency of the LF (Low Frequency) component based on the frequency analysis of the heartbeat variation, and information related to the standard deviation of the Lorentz curve of the heartbeat variation.

[0173] (Note 13) According to the information processing system described in Note 9, wherein, The pulsation information includes: the interquartile range of the sample entropy of the heart rate fluctuations; the interquartile range of the standard deviation of the heart rate obtained from the electrocardiogram information; the median of the value obtained by dividing the standard deviation of the Lorenz curve of the heart rate fluctuations along the y=x axis by the standard deviation of the Lorenz curve along the y=-x axis; the interquartile range of the value obtained by dividing the standard deviation of the Lorenz curve of the heart rate fluctuations along the y=x axis by the standard deviation of the Lorenz curve along the y=-x axis; the median of the logarithm of the product of the standard deviation of the Lorenz curve of the heart rate fluctuations along the y=x axis and the standard deviation of the Lorenz curve along the y=-x axis; and the LF (Low Frequency) analysis based on the heart rate fluctuations. The interquartile range of the frequency (low frequency) correction value, the median of the peak frequency of the LF component analyzed by frequency resolution of the heart rate variation, the interquartile range of the mean of the heart rate variation, the interquartile range of the standard deviation of the Lorentz curve of the heart rate variation along the y=x axis, the interquartile range of the peak frequency of the HF (High Frequency) component analyzed by frequency resolution of the heart rate variation, the median of the integral value of the LF component analyzed by frequency resolution of the heart rate variation, the interquartile range of the approximate entropy of the heart rate variation, the median of the logarithm of the HF component analyzed by frequency resolution of the heart rate variation, and the median of the LF correction value analyzed by frequency resolution of the heart rate variation, at least one of the following:

[0174] (Note 14) According to the information processing system described in Note 13, wherein, The frequency analysis includes at least one of high-speed Fourier transform, Lomb-Scargle Periodogram, and autoregressive model.

[0175] (Note 15) According to the information processing system described in Note 5, wherein, The pulsation information is different from the electrocardiogram waveform contained in the electrocardiogram information.

[0176] (Note 16) According to the information processing system described in Note 5, wherein, The acquisition unit calculates the pulsation information based on the electrocardiogram information during the sleep period in the time-series electrocardiogram information.

[0177] (Note 17) According to the information processing system described in Note 16, wherein, The acquisition unit determines the sleep period based on at least one of a preset time, temporal electrocardiogram information detected from the subject, temporal acceleration information detected from the subject, and temporal posture information detected from the subject.

[0178] (Note 18) According to the information processing system described in Note 16, wherein, The electrocardiogram (ECG) information during the sleep period corresponds to the ECG information from 0:00 to 6:00 in the time sequence ECG information.

[0179] (Note 19) According to the information processing system described in Note 5, wherein, It also has a detection unit installed on the object and for detecting the electrocardiogram information.

[0180] (Note 20) According to the information processing system described in Note 19, wherein, The detection unit is installed on the torso of the subject.

[0181] (Note 21) According to the information processing system described in Note 19, wherein, The detection unit is a flexible plate-shaped part that is attached to or adsorbed onto the skin of the subject's torso.

[0182] (Note 22) According to the information processing system described in Note 1, wherein, The determination result indicates information about the presence and severity of heart failure in the subject. The diagnostic information is information representing the cardiac function status associated with the provider's heart failure.

[0183] (Note 23) A program in which, The computer performs a judgment process, which involves inputting pulsation information related to the subject's heartbeat and the subject's attribute information into the learned model to generate a judgment result indicating the presence and severity of the subject's heart disease. The learned model is a learned model obtained by learning from multiple providers who provide training data, using pulsation information related to the heartbeat of the provider, attribute information of the provider, and diagnostic information representing the cardiac function status of the provider.

[0184] (Appendix 24) An information processing device, wherein, It has a determination unit that, by inputting pulsation information related to the subject's heartbeat and the subject's attribute information into the learned model, generates a determination result indicating the presence and severity of the subject's heart disease. The learned model is a learned model obtained by learning from multiple providers who provide training data, using pulsation information related to the heartbeat of the provider, attribute information of the provider, and diagnostic information representing the cardiac function status of the provider.

[0185] (Appendix 25) An information processing device, wherein, It has a study department. In order to generate a learned model that outputs a judgment result indicating the presence and severity of heart disease in a subject by inputting pulsation information related to the subject's heartbeat and attribute information of the subject, the learning unit performs machine learning for multiple providers who provide training data, using pulsation information related to the heartbeat of the provider, attribute information of the provider, and diagnostic information indicating the cardiac function status of the provider.

[0186] Explanation of reference numerals in the attached figures

[0187] 1, 2: Information Processing System

[0188] 10, 40: Information processing device

[0189] 11: CPU

[0190] 12: ROM

[0191] 13: RAM

[0192] 14: Auxiliary storage device

[0193] 15: Input device

[0194] 16: Display device

[0195] 17: External I / F

[0196] 20: Wearable sensors

[0197] 30: Operating terminal

[0198] 101: Acquisition Department

[0199] 102: Judgment Department

[0200] 103: Output Control Unit

[0201] 104: Study Department.

Claims

1. An information processing system, wherein, It has a determination unit that, by inputting pulsation information related to the subject's heartbeat and the subject's attribute information into the learned model, generates a determination result indicating the presence and severity of the subject's heart disease. The learned model is a learned model obtained by learning from multiple providers who provide training data, using pulsation information related to the heartbeat of the provider, attribute information of the provider, and diagnostic information representing the cardiac function status of the provider.

2. The information processing system according to claim 1, wherein, The providers include healthy individuals and patients diagnosed with heart failure by a doctor.

3. The information processing system according to claim 1, wherein, It also includes a unit for acquiring the pulsation information of the subject.

4. The information processing system according to claim 1, wherein, The attribute information of the subject includes at least one of the subject's age, gender, height, weight, and BMI (Body Mass Index).

5. The information processing system according to claim 3, wherein, The acquisition unit calculates the pulsation information based on the temporal electrocardiogram information detected from the subject.

6. The information processing system according to claim 5, wherein, The acquisition unit performs the following actions: By segmenting the electrocardiogram (ECG) information at predetermined intervals along the temporal direction, a plurality of segmented data are generated in a temporal sequence. For each of the multiple segmented data, calculate a multiple number of statistical values. For each of the multiple statistical values, the pulsation information representing the temporal direction characteristics of the electrocardiogram information is calculated.

7. The information processing system according to claim 5, wherein, The electrocardiogram information includes more than 72 hours of continuous data obtained from the subject.

8. The information processing system according to claim 5, wherein, The electrocardiogram information includes more than 120 hours of continuous data obtained from the subject.

9. The information processing system according to claim 6, wherein, The statistical values ​​include at least one of the following: various statistical values ​​of heart rate variability obtained from the electrocardiogram information, the peak frequency of a specified frequency band component in the power spectrum based on the frequency analysis of the heart rate variability, the integral value of the specified frequency band component, and the value based on the nonlinear analysis of the heart rate variability. The pulsation information includes at least one of the median and interquartile range of the statistical values.

10. The information processing system according to claim 9, wherein, The attribute information of the subject is the subject's age and BMI (Body Mass Index). The pulsation information is the interquartile range of the sample entropy of the heartbeat variation.

11. The information processing system according to claim 10, wherein, The pulsation information further comprises: the median of the LF (Low Frequency) correction value based on the frequency analysis of the heart rate variation, the interquartile range of the standard deviation of the heart rate obtained from the electrocardiogram information, the median of the peak frequency of the LF component based on the frequency analysis of the heart rate variation, the median of the logarithm of the HF (High Frequency) component based on the frequency analysis of the heart rate variation, and the interquartile range of the LF correction value based on the frequency analysis of the heart rate variation.

12. The information processing system according to claim 9, wherein, The pulsation information includes at least one of the following: information related to the entropy of the heartbeat variation, the average heartbeat interval in the heartbeat variation, the heart rate based on the electrocardiogram information, information related to the integral value or peak frequency of the HF (High Frequency) component based on the frequency analysis of the heartbeat variation, information related to the integral value or peak frequency of the LF (Low Frequency) component based on the frequency analysis of the heartbeat variation, and information related to the standard deviation of the Lorenz curve of the heartbeat variation.

13. The information processing system according to claim 9, wherein, The pulsation information includes: the interquartile range of the sample entropy of the heart rate fluctuations; the interquartile range of the standard deviation of the heart rate obtained from the electrocardiogram information; the median of the value obtained by dividing the standard deviation of the Lorenz curve of the heart rate fluctuations along the y=x axis by the standard deviation of the Lorenz curve along the y=-x axis; the interquartile range of the value obtained by dividing the standard deviation of the Lorenz curve of the heart rate fluctuations along the y=x axis by the standard deviation of the Lorenz curve along the y=-x axis; the median of the logarithm of the product of the standard deviation of the Lorenz curve of the heart rate fluctuations along the y=x axis and the standard deviation of the Lorenz curve along the y=-x axis; and the LF (Low Frequency) analysis based on the heart rate fluctuations. The interquartile range of the frequency (low frequency) correction value, the median of the peak frequency of the LF component based on the frequency analysis of the heart rate variation, the interquartile range of the mean of the heart rate variation, the interquartile range of the standard deviation of the Lorentz curve of the heart rate variation along the y=x axis, the interquartile range of the peak frequency of the HF (High Frequency) component based on the frequency analysis of the heart rate variation, the median of the integral value of the LF component based on the frequency analysis of the heart rate variation, the interquartile range of the approximate entropy of the heart rate variation, the median of the logarithm of the HF component based on the frequency analysis of the heart rate variation, and the median of the LF correction value based on the frequency analysis of the heart rate variation.

14. The information processing system according to claim 13, wherein, The frequency analysis includes at least one of the following: high-speed Fourier transform, Lomb-Scargle Periodogram, and autoregressive model.

15. The information processing system according to claim 5, wherein, The pulsation information is different from the electrocardiogram waveform contained in the electrocardiogram information.

16. The information processing system according to claim 5, wherein, The acquisition unit calculates the pulsation information based on the electrocardiogram information during the sleep period in the time-series electrocardiogram information.

17. The information processing system according to claim 16, wherein, The acquisition unit determines the sleep period based on at least one of a preset time, temporal electrocardiogram information detected from the subject, temporal acceleration information detected from the subject, and temporal posture information detected from the subject.

18. The information processing system according to claim 16, wherein, The electrocardiogram (ECG) information during the sleep period corresponds to the ECG information from 0:00 to 6:00 in the time sequence ECG information.

19. The information processing system according to claim 5, wherein, It also includes a detection unit installed on the object and detecting the electrocardiogram information.

20. The information processing system according to claim 19, wherein, The detection unit is installed on the torso of the subject.

21. The information processing system according to claim 19, wherein, The detection unit is a flexible plate-shaped part that is attached to or adsorbed onto the skin of the subject's torso.

22. The information processing system according to claim 1, wherein, The determination result indicates information about the presence and severity of heart failure in the subject. The diagnostic information is information representing the cardiac function status associated with the provider's heart failure.

23. A program in which, The computer performs a judgment process, which involves inputting pulsation information related to the subject's heartbeat and the subject's attribute information into the learned model to generate a judgment result indicating the presence and severity of the subject's heart disease. The learned model is a learned model obtained by learning from multiple providers who provide training data, using pulsation information related to the heartbeat of the provider, attribute information of the provider, and diagnostic information representing the cardiac function status of the provider.

24. An information processing apparatus, wherein, It has a determination unit that, by inputting pulsation information related to the subject's heartbeat and the subject's attribute information into the learned model, generates a determination result indicating the presence and severity of the subject's heart disease. The learned model is a learned model obtained by learning from multiple providers who provide training data, using pulsation information related to the heartbeat of the provider, attribute information of the provider, and diagnostic information representing the cardiac function status of the provider.

25. An information processing apparatus, wherein, The system includes a learning unit that, in order to generate a learned model that outputs a judgment result indicating the presence and severity of heart disease in a subject by inputting pulsation information related to the subject's heartbeat and attribute information of the subject, performs machine learning on multiple providers of training data using pulsation information related to the heartbeat of the providers, attribute information of the providers, and diagnostic information indicating the cardiac function status of the providers.

Images