Classification device, classification method, program, recording medium, and diagnosis assistance device for evaluating autonomic dysfunction

Abstract

An acquisition unit (312) acquires RRI data. A generation unit (314) generates heart rate variability index data on the basis of the RRI data. A first classification unit (318) classifies the heart rate variability index data into a combination of first and second groups, and a third group by inputting the heart rate variability index data to a first algorithm. A second classification unit (320) classifies the heart rate variability index data classified into the combination of first and second groups into the first group and the second group. The first algorithm is trained by using teaching data for the combination of first and second groups, and teaching data for the third group.

Description

Classification device, classification method, program, recording medium, and diagnostic support device for evaluating autonomic nervous dysfunction

[0001] The present disclosure relates to a classification technique, and in particular to a classification device, classification method, program, recording medium, and diagnostic support device for evaluating autonomic nervous dysfunction that use RR interval (RRI) data.

[0002] Orthostatic hypotension (OH) is a typical autonomic dysfunction observed in patients with idiopathic REM behavior disorder (iRBD). The orthostatic challenge test to determine the presence or absence of OH is a standard autonomic function test that measures the drop in blood pressure when changing posture from supine to standing. However, the orthostatic challenge test poses a risk of patient fall when standing and requires several medical professionals to surround the patient during the test, placing a significant burden on both the patient and the medical professionals. Therefore, a simple method for assessing autonomic function is needed. Given the presence of autonomic dysfunction in OH patients, a correlation between heart rate variability (HRV) in the supine position and the occurrence of OH during an orthostatic stress test has been reported (see, for example, Non-Patent Document 1).

[0003] Yukiyoshi Sumi, Chikao Nakayama, Hiroshi Kadotani, Masahiro Matsuo, Yuji Ozeki, Takafumi Kinoshita, Yuki Goto, Manabu Kano, Toshitaka Yamakawa, Masako Hasegawa-Ohira, Keiko Ogawa and Koichi Fujiwara, “Resting Heart Rate Variability Is Associated With Subsequent Orthostatic Hypotension: Comparison Between Healthy Older People and Patients With Rapid Eye "Movement Behavior Sleep Disorder", Frontiers in Neurology, November 2020, Volume 11, Article 567984, p. 1-15

[0004] It is desired to more simply determine the presence or absence of OH using a heart rate variability index.

[0005] The present disclosure has been made in light of these circumstances, and aims to provide a technique for classifying the condition of a subject using a heart rate variability index.

[0006] In order to solve the above problem, a classification device according to one aspect of the present disclosure includes an acquisition unit that acquires RRI data indicating the interval between R waves in an electrocardiogram signal of a subject, a generation unit that generates heart rate variability index data including index values ​​for each of a plurality of types of heart rate-related indexes based on the RRI data acquired by the acquisition unit, a first classification unit that classifies the heart rate variability index data into a combination of a first group and a second group and a third group by inputting the heart rate variability index data generated by the generation unit into a first algorithm, and a second classification unit that classifies the heart rate variability index data classified into a combination of the first group and the second group by the first classification unit into the first group and the second group. The first algorithm is trained using teacher data for the combination of the first group and the second group and teacher data for the third group.

[0007] Another aspect of the present disclosure is a classification method. The method includes the steps of acquiring RRI data indicating the interval between R waves in an electrocardiogram signal of a subject, generating heart rate variability index data including index values ​​for multiple types of heart rate-related indexes based on the acquired RRI data, inputting the generated heart rate variability index data into a first algorithm to classify the heart rate variability index data into a combination of a first group and a second group and a third group, and classifying the heart rate variability index data classified into the combination of the first group and the second group into the first group and the second group. The first algorithm is trained using teacher data for the combination of the first group and the second group and teacher data for the third group.

[0008] Yet another aspect of the present disclosure is a diagnostic support device. The device includes: an acquisition unit that acquires RRI data indicating the interval between R waves in an electrocardiogram signal of a subject; a generation unit that generates heart rate variability index data including index values ​​for multiple types of heart rate-related indexes based on the RRI data acquired by the acquisition unit; a first classification unit that inputs the heart rate variability index data generated by the generation unit into a first algorithm to classify the heart rate variability index data into a combination of a first group and a second group and a third group; and a second classification unit that classifies the heart rate variability index data classified into a combination of the first group and the second group by the first classification unit into the first group and the second group. The first algorithm evaluates an autonomic nervous dysfunction that has been trained using teacher data for the combination of the first group and the second group and teacher data for the third group.

[0009] Any combination of the above components, and conversion of the present disclosure into a method, device, system, recording medium, computer program, etc., are also valid aspects of the present disclosure.

[0010] According to the present disclosure, the heart rate variability index can be used to classify the condition of a subject.

[0011] FIG. 1 is a diagram showing a comparison of the heart rate variability analysis results in Non-Patent Document 1 with the HC group, the OH(-)iRBD group, and the OH(+)iRBD group. FIG. 2 is a diagram showing the configuration of a classification system according to the present embodiment. FIG. 3 is a diagram showing the configuration of the information processing device of FIG. 1. FIG. 4 is a diagram showing R waves contained in a signal received by the communication unit of FIG. 3. FIG. 5 is a diagram showing evaluation results of classification performance by the information processing device of FIG. 3. FIG. 6 is a diagram showing a confusion matrix in the information processing device of FIG. 3.

[0012] Before describing the present disclosure in detail, an overview will be provided. This example relates to a classification system that classifies subjects into a healthy group, an OH(-) iRBD group, and an OH(+) iRBD group based on a heart rate variability (HRV) index generated from a subject's RRI data. As mentioned above, a simple method for evaluating autonomic nervous function is needed. In light of reports that HRV differs between healthy elderly people and iRBD patients, this example classifies subjects into three groups: healthy subjects, iRBD patients without OH, and iRBD patients with OH, by running a machine learning (ML) model based on HRV in the supine position during an orthostatic stress test.

[0013] To clearly explain this embodiment, the results of the heart rate variability analysis in Non-Patent Document 1 will be described first. In Non-Patent Document 1, electrocardiograms were measured during an orthostatic stress test in which subjects maintained a supine position and an upright position for 15 minutes each. The subjects were also divided into three groups: healthy subjects (HC), iRBD patients without OH symptoms (OH(-)iRBD), and iRBD patients with OH symptoms (OH(+)iRBD). In this situation, Non-Patent Document 1 compared HRV among the three groups (HC, OH(-)iRBD, and OH(+)iRBD) during the five minutes of supine positioning immediately prior to standing. Specifically, HRV measured in the supine position during the test was calculated using time domain analysis and Poincaré plots to evaluate autonomic dysfunction.

[0014] Furthermore, HRV index data is data that includes index values ​​for multiple types of heart rate-related indexes and reflects autonomic nervous activity. HRV refers to millisecond-order fluctuations in the ECG RRI. When the heart beats quickly, the RRI becomes shorter. RRI data is generated only when R waves occur, so the sampling interval is not constant. HRV is generated by the balance between sympathetic and parasympathetic nervous activity. When the RRI increases, roughly speaking, the parasympathetic nervous system is dominant and the person is in a relaxed state, while when the RRI decreases, the sympathetic nervous system is dominant and the person is in a stressed state. Therefore, the RRI can be treated as a stress index.

[0015] HRV index data refers to several values ​​that serve as indicators for evaluating HRV. HRV index data can be calculated from the RRI, and are classified into time domain indexes calculated from the RRI on the time axis, frequency domain indexes calculated from the frequency density of the RRI, and Poincaré domain indexes calculated from a Poincaré plot. Typically, a time window of 2 minutes or more is used for HRV analysis, and a 5-minute time window is used in this embodiment.

[0016] The time domain indexes include, for example, SDNN, RMSSD, and pNN50, and the Poincaré domain indexes include, for example, SD1, SD2, and SD1 / SD2. The time domain HRV index data is calculated directly from the RRI data. Among the time domain HRV index data, SDNN is the standard deviation of the RRI data included in the time window. RMSSD is the root mean square of the difference between two temporally adjacent RRI data included in the time window. pNN50 is a value obtained by dividing NN50 by the total number of RRI data within the time window.

[0017] A Poincaré plot is a plot that expresses the characteristics of RRI two-dimensionally. In a Poincaré plot, the x-axis plots the value of RRI(n) and the x-axis plots RRI(n+1). n indicates the ordinal number of the RRI data. From such a Poincaré plot, Poincaré zone indices are calculated. The Poincaré zone indices include SD1 and SD2. SD1 is the standard deviation in the vertical direction of the Poincaré plot, and SD2 is the standard deviation in the horizontal direction of the Poincaré plot. SD1 / SD2 is the ratio of SD1 to SD2.

[0018] Figure 1 shows the results of the heart rate variability analysis in Non-Patent Document 1 and a comparison of the HC group, OH(-)iRBD group, and OH(+)iRBD group. Here, there were 42 HCs, 12 OH(-)iRBDs, and 9 OH(+)iRBDs. From this, significant differences were observed between the HC group, OH(-)iRBD group, and OH(+)iRBD group in SDNN, RMSSD, pNN50, SD1, and SD2. In particular, the HRV of the HC group was significantly different from the HRV of the OH(-)iRBD group and the HRV of the OH(+)iRBD group.

[0019] FIG. 2 shows the configuration of the classification system 1000. The classification system 1000 performs a learning operation to learn an algorithm for classifying into the HC group, the OH(-)iRBD group, and the OH(+)iRBD group, and a classification operation to perform classification using the learned algorithm. FIG. 2 corresponds to the configuration of the classification system 1000 for performing the classification operation. The classification system 1000 includes a plurality of electrodes 100, a heartbeat sensor 200, and an information processing device 300.

[0020] The subject 10 is a person to be classified into the HC group, the OH(-)iRBD group, and the OH(+)iRBD group. Three electrodes 100 are attached to the body surface of the subject 10. The three electrodes 100 are composed of, for example, a positive electrode, a negative electrode, and a ground electrode. The three electrodes 100 are connected to a heart rate sensor 200. The heart rate sensor 200 extracts R waves from an electrocardiogram signal measured for the subject 10 via the three electrodes 100. For example, the electrocardiogram signal of the subject 10 is measured for 30 seconds to 10 minutes in a supine position, and R waves are extracted on the order of milliseconds. The heart rate sensor 200 also has a wireless communication function such as Bluetooth (registered trademark) LE and wirelessly transmits a signal indicating the extracted R waves. The configuration of the electrodes 100 and the heart rate sensor 200 is not limited to that shown in FIG. 2 .

[0021] The information processing device 300 is, for example, a computer or a smartphone. The information processing device 300 has a wireless communication function such as Bluetooth (registered trademark) LE, and receives a signal indicating an R wave wirelessly transmitted from the heart rate sensor 200. The information processing device 300 stores a learned algorithm in advance, and classifies the subject 10 into one of the HC group, the OH(-)iRBD group, and the OH(+)iRBD group using the learned algorithm.

[0022] In the learning operation, three electrodes 100 are attached to the body surface of a donor (not shown) instead of the subject 10. The donor is a person who provides electrocardiogram signals necessary for learning the algorithm. There may be multiple donors. The information processing device 300 learns the algorithm based on signals indicating the R waves of the donor. The information processing device 300 used in the learning operation and the information processing device 300 used in the classification operation may be different devices.

[0023] FIG. 3 shows the configuration of an information processing device 300. The information processing device 300 includes a communication unit 310, an acquisition unit 312, a generation unit 314, a storage unit 316, a first classification unit 318, a second classification unit 320, and an output unit 322. The communication unit 310 receives a signal indicating an R wave wirelessly transmitted from the heartbeat sensor 200. FIG. 4 shows an R wave included in the signal received by the communication unit 310. The potential distribution generated by the electrical activity of the heart is detected as a potential difference between two points of electrodes 100 placed on the body surface or inside the body. The waveform representing the potential difference between the two points that fluctuates over time is called an ECG. Periodic potential changes appear in the ECG, and several peaks can be seen, the highest of which is called an R wave. Returning to the explanation of FIG. 3 .

[0024] The communication unit 310 generates data indicating R waves (hereinafter referred to as "R wave data") from the received signal and outputs the data to the acquisition unit 312. The R wave data is digital data that is, for example, "1" when the amplitude of the signal indicating the R wave is equal to or greater than a threshold voltage, and is "0" when the amplitude is less than the threshold voltage. In other words, the R wave data is a rectangular pulse train in which periods corresponding to R waves in the electrocardiogram signal are set to "1" and other periods are set to "0."

[0025] The acquisition unit 312 receives R-wave data from the communication unit 310. Based on the R-wave data, the acquisition unit 312 acquires RRI data, which is time-series data of an RRI variable indicating the interval between R waves. The RRI data indicates the interval between R waves in the electrocardiogram signal of the subject 10. In this case, the acquisition unit 312 calculates the time interval between the falling edges of two temporally adjacent rectangular pulses as an RRI variable from the R-wave data, and generates the RRI data by arranging the calculated RRI variables in a time series. Specifically, the acquisition unit 312 first calculates, for the R-wave data, the time interval between a first time when an R wave is detected and a second time when the immediately preceding R wave is detected as the RRI value at the first time. This generates data indicating a change in the RRI value over time. The acquisition unit 312 then performs spline interpolation on the data indicating the time-dependent changes in the RRI values ​​and samples them at equal intervals to generate RRI data indicating the time-dependent changes in the RRI values ​​arranged at equal intervals. In other words, the acquisition unit 312 performs spline interpolation (interpolation processing) on ​​the time-dependent data obtained from the electrocardiogram signal at different time intervals and samples them at equal intervals to generate RRI data arranged at equal intervals. Here, the RRI data acquired by the acquisition unit 312 is data from 30 seconds to 10 minutes when the subject 10 is in the supine position. The acquisition unit 312 outputs the RRI data to the generation unit 314.

[0026] The generating unit 314 generates heart rate variability (HRV) index data including index values ​​for multiple types of heart rate-related indexes based on the RRI data acquired by the acquiring unit 312. The HRV index data is as described above, and includes, for example, SDNN, RMSSD, pNN50, SD1, SD2, and SD1 / SD2. The HRV index data is not limited to these. The generating unit 314 outputs the generated HRV index data to the first classifying unit 318.

[0027] The storage unit 316 stores the first algorithm and the second algorithm learned in the learning operation. Specific examples of the first algorithm and the second algorithm will be described later. The first algorithm is an algorithm for classifying HRV index data into an iRBD group and an HC group. The iRBD group is a combination of the OH(-)iRBD group and the OH(+)iRBD group, i.e., patients with idiopathic REM sleep behavior disorder. The first algorithm is learned in the learning operation using training data for the iRBD group and training data for the HC group.

[0028] The second algorithm is an algorithm for classifying the HRV index data of the iRBD group into an OH(-)iRBD group and an OH(+)iRBD group. The OH(-)iRBD group is made up of idiopathic REM sleep behavior disorder patients without orthostatic hypotension symptoms, and the OH(+)iRBD group is made up of idiopathic REM sleep behavior disorder patients with orthostatic hypotension symptoms. The second algorithm is trained using training data for the OH(-)iRBD group and training data for the OH(+)iRBD group in the training operation.

[0029] The first classification unit 318 acquires the first algorithm from the storage unit 316. The first classification unit 318 inputs the HRV index data generated by the generation unit 314 into the first algorithm, thereby classifying the HRV index data into an iRBD group and an HC group. That is, the first classification unit 318 classifies the HRV index data into the iRBD group and the HC group, which have significantly different HRVs, as shown in FIG. 1 . The first classification unit 318 outputs the HRV index data classified into the iRBD group to the second classification unit 320. The first classification unit 318 also outputs the HRV index data classified into the HC group to the output unit 322.

[0030] The second classifying unit 320 obtains the second algorithm from the storage unit 316. The second classifying unit 320 inputs the HRV index data classified into the iRBD group by the first classifying unit 318 into the second algorithm, thereby classifying the HRV index data classified into the iRBD group into an OH(-) iRBD group and an OH(+) iRBD group. The second classifying unit 320 outputs the HRV index data classified into the OH(-) iRBD group and the HRV index data classified into the OH(+) iRBD group to the output unit 322.

[0031] The output unit 322 receives the HRV index data classified into the HC group from the first classification unit 318. The output unit 322 receives the HRV index data classified into the OH(-)iRBD group and the HRV index data classified into the OH(+)iRBD group from the second classification unit 320. The output unit 322 outputs the classification results.

[0032] Specific examples of the first and second algorithms are described below. In this embodiment, XGBoost (xgb), LightGBM (lgb), Random Forest (Rf), Support Vector Machine (SVM), and Logistic Regression (log) are used as candidates for the first and second algorithms. FIG. 5 shows the results of evaluation of classification performance by the information processing device 300. Here, classification performance was evaluated for all combinations of five types of candidate algorithms for the first and second algorithms. Classification performance is evaluated using Accuracy and F-measure.

[0033] FIG. 6 is a diagram showing a confusion matrix in the information processing device 300. Here, random forest is used as the first algorithm, and logistic regression is used as the second algorithm. In the confusion matrix, vertical labels indicate true values, and horizontal labels indicate predicted values ​​by the algorithm. Furthermore, "0" indicates the HC group, "1" indicates the OH(-)iRBD group, and "2" indicates the OH(+)iRBD group. The numbers in the squares indicate the number of samples that fit that combination. Cells where the value of the vertical label and the value of the horizontal label are the same indicate data where the correct answer and prediction match, and other cells indicate data where there is no match. This indicates that the classification accuracy using the first algorithm is high.

[0034] This configuration can be realized in hardware terms by any computer's CPU (Central Processing Unit), memory, and other LSIs (Large Scale Integration), and in software terms by programs loaded into memory, but here we depict functional blocks realized by the cooperation of these. Therefore, those skilled in the art will understand that these functional blocks can be realized in various forms by hardware alone or a combination of hardware and software.

[0035] According to this embodiment, the HRV index data is classified into a combination of Group 1 and Group 2 and Group 3 using a first algorithm, and the combination of Group 1 and Group 2 is classified into Group 1 and Group 2 using a second algorithm, so that the condition of the subject can be classified using the HRV index data. Also, OH(-)iRBD, OH(+)iRBD, and HC can be classified using the HRV index data. Furthermore, since the HRV index data is used to classify OH(-)iRBD, OH(+)iRBD, and HC, an orthostatic stress test can be eliminated. Furthermore, since an orthostatic stress test is not required, the risk of injury to the patient can be reduced. Furthermore, since an orthostatic stress test is not required, the burden on medical professionals accompanying the patient during the orthostatic stress test can be reduced.

[0036] In order to solve the above problem, a classification device according to one aspect of the present disclosure includes an acquisition unit that acquires RRI data indicating the interval between R waves in an electrocardiogram signal of a subject, a generation unit that generates heart rate variability index data including index values ​​for each of a plurality of types of heart rate-related indexes based on the RRI data acquired by the acquisition unit, a first classification unit that classifies the heart rate variability index data into a combination of a first group and a second group and a third group by inputting the heart rate variability index data generated by the generation unit into a first algorithm, and a second classification unit that classifies the heart rate variability index data classified into a combination of the first group and the second group by the first classification unit into the first group and the second group. The first algorithm is trained using teacher data for the combination of the first group and the second group and teacher data for the third group.

[0037] According to this aspect, the heart rate variability index data is classified into a combination of the first group and the second group and a third group using a first algorithm, and the combination of the first group and the second group is classified into the first group and the second group, so that the condition of the subject can be classified using the heart rate variability index.

[0038] The second classification unit classifies the heart rate variability index data classified into a combination of the first and second groups into the first and second groups by inputting the heart rate variability index data classified into a combination of the first and second groups by the first classification unit into the second algorithm, and the second algorithm may be trained using teacher data for the first group and teacher data for the second group. In this case, the combination of the first and second groups is classified into the first and second groups by the second algorithm, so that the condition of the subject can be classified using the heart rate variability index.

[0039] The first group may be idiopathic REM sleep behavior disorder patients without orthostatic hypotension symptoms, the second group may be idiopathic REM sleep behavior disorder patients with orthostatic hypotension symptoms, a combination of the first and second groups may be idiopathic REM sleep behavior disorder patients, and the third group may be healthy individuals. In this case, the heart rate variability index can be used to classify the idiopathic REM sleep behavior disorder patients without orthostatic hypotension symptoms, the idiopathic REM sleep behavior disorder patients with orthostatic hypotension symptoms, and the healthy individuals.

[0040] The first algorithm and the second algorithm may be different algorithms.

[0041] The first algorithm may be a random forest and the second algorithm may be a logistic regression.

[0042] The RRI data acquired by the acquisition unit may be data for 30 seconds to 10 minutes when the subject is in a supine position.

[0043] The heart rate variability index data generated by the generation unit may include SDNN, RMSSD, pNN50, SD1, SD2, and SD1 / SD2.

[0044] Another aspect of the present disclosure is a classification method. The method includes the steps of acquiring RRI data indicating the interval between R waves in an electrocardiogram signal of a subject, generating heart rate variability index data including index values ​​for multiple types of heart rate-related indexes based on the acquired RRI data, inputting the generated heart rate variability index data into a first algorithm to classify the heart rate variability index data into a combination of a first group and a second group and a third group, and classifying the heart rate variability index data classified into the combination of the first group and the second group into the first group and the second group. The first algorithm is trained using teacher data for the combination of the first group and the second group and teacher data for the third group.

[0045] Yet another aspect of the present disclosure is a diagnostic support device. The device includes: an acquisition unit that acquires RRI data indicating the interval between R waves in an electrocardiogram signal of a subject; a generation unit that generates heart rate variability index data including index values ​​for multiple types of heart rate-related indexes based on the RRI data acquired by the acquisition unit; a first classification unit that inputs the heart rate variability index data generated by the generation unit into a first algorithm to classify the heart rate variability index data into a combination of a first group and a second group and a third group; and a second classification unit that classifies the heart rate variability index data classified into a combination of the first group and the second group by the first classification unit into the first group and the second group. The first algorithm evaluates an autonomic nervous dysfunction that has been trained using teacher data for the combination of the first group and the second group and teacher data for the third group.

[0046] The present disclosure has been described above based on examples. These examples are merely illustrative, and it will be understood by those skilled in the art that various modifications are possible in the combination of the components and processing steps, and that such modifications are also within the scope of the present disclosure.

[0047] According to the present disclosure, the heart rate variability index can be used to classify the condition of a subject.

[0048] 10 Subject, 100 Electrode, 200 Heartbeat sensor, 300 Information processing device, 310 Communication unit, 312 Acquisition unit, 314 Generation unit, 316 Storage unit, 318 First classification unit, 320 Second classification unit, 322 Output unit, 1000 Classification system.

Claims

1. A classification device comprising: an acquisition unit that acquires RRI data indicating the interval between R waves in a subject's electrocardiogram signal; a generation unit that generates heart rate variability index data including index values ​​for each of a plurality of types of indices related to heart rate based on the RRI data acquired by the acquisition unit; a first classification unit that classifies the heart rate variability index data into a combination of a first group and a second group and a third group by inputting the heart rate variability index data generated by the generation unit into a first algorithm; and a second classification unit that classifies the heart rate variability index data classified into the combination of the first group and the second group by the first classification unit into the first group and the second group, wherein the first algorithm is trained using teacher data for the combination of the first group and the second group and teacher data for the third group.

2. The classification device described in claim 1, wherein the second classification unit classifies the heart rate variability index data classified into a combination of the first group and the second group into the first group and the second group by inputting the heart rate variability index data classified into a combination of the first group and the second group in the first classification unit into a second algorithm, and the second algorithm is trained using teacher data for the first group and teacher data for the second group.

3. The classification device of claim 2, wherein the first group is idiopathic REM sleep behavior disorder patients without orthostatic hypotension symptoms, the second group is idiopathic REM sleep behavior disorder patients with orthostatic hypotension symptoms, a combination of the first group and the second group is idiopathic REM sleep behavior disorder patients, and the third group is healthy individuals.

4. The classification device according to claim 2 or 3, wherein the first algorithm and the second algorithm are different algorithms.

5. The classification device of claim 4, wherein the first algorithm is a random forest and the second algorithm is a logistic regression.

6. A classification device according to any one of claims 1 to 5, wherein the RRI data acquired by the acquisition unit is data for 30 seconds to 10 minutes when the subject is in a supine position.

7. A classification device according to any one of claims 1 to 6, wherein the heart rate variability index data generated by the generation unit includes SDNN, RMSSD, pNN50, SD1, SD2, and SD1 / SD2.

8. A classification method comprising the steps of: acquiring RRI data indicating the interval between R waves in a subject's electrocardiogram signal; generating heart rate variability index data including index values ​​for each of a plurality of types of indices related to heart rate based on the acquired RRI data; classifying the heart rate variability index data into a combination of a first group and a second group and a third group by inputting the generated heart rate variability index data into a first algorithm; and classifying the heart rate variability index data classified into the combination of the first group and the second group into the first group and the second group, wherein the first algorithm is trained using teacher data for the combination of the first group and the second group and teacher data for the third group.

9. A program for causing a computer to execute the steps of: acquiring RRI data indicating the interval between R waves in a subject's electrocardiogram signal; generating heart rate variability index data including index values ​​for each of a plurality of types of indices related to heart rate based on the acquired RRI data; classifying the heart rate variability index data into a combination of a first group and a second group and a third group by inputting the generated heart rate variability index data into a first algorithm; and classifying the heart rate variability index data classified into a combination of the first group and the second group into the first group and the second group, wherein the first algorithm is learned using teacher data for the combination of the first group and the second group and teacher data for the third group.

10. A recording medium storing a program for causing a computer to execute a method comprising the steps of: acquiring RRI data indicating the interval between R waves in a subject's electrocardiogram signal; generating heart rate variability index data including index values ​​for each of a plurality of types of indices related to heart rate based on the acquired RRI data; classifying the heart rate variability index data into a combination of a first group and a second group and a third group by inputting the generated heart rate variability index data into a first algorithm; and classifying the heart rate variability index data classified into a combination of the first group and the second group into the first group and the second group, wherein the first algorithm is learned using teacher data for the combination of the first group and the second group and teacher data for the third group.

11. A diagnostic support device for evaluating autonomic nervous dysfunction comprising: an acquisition unit that acquires RRI data indicating the interval between R waves in an electrocardiogram signal of a subject; a generation unit that generates heart rate variability index data including index values ​​for each of a plurality of types of indices related to heart rate based on the RRI data acquired by the acquisition unit; a first classification unit that classifies the heart rate variability index data into a combination of a first group and a second group and a third group by inputting the heart rate variability index data generated by the generation unit into a first algorithm; and a second classification unit that classifies the heart rate variability index data classified into the combination of the first group and the second group by the first classification unit into the first group and the second group, wherein the first algorithm is learned using teacher data for the combination of the first group and the second group and teacher data for the third group.

Images