JP2020151358A - Sleep apnea syndrome determination apparatus, sleep apnea syndrome determination method, and sleep apnea syndrome determination program - Google Patents

Abstract

To enable the determination of sleep apnea syndrome without imposing a burden on a subject.SOLUTION: Biological vibration data is obtained on the basis of body movement or pressure change during sleep of a subject. A plurality of feature quantities or frequency spectra are then calculated at a regular interval from the acquired biological vibration data. Furthermore, after determining sleep stages from the plurality of calculated feature quantities or frequency spectra using a plurality of decision trees having branching conditions for determining the sleep stages, a sleep apnea syndrome is determined on the basis of a determination result obtained by the combination of the classification criteria of the decision trees used for the determination. Alternatively, a power spectrum of a frequency component included in the biological vibration data is calculated, then the contribution degrees to the classification of the plurality of decision trees each having different branching conditions are calculated with respect to the power spectrum of the frequency component, and a sleep apnea syndrome is determined on the basis of the calculated contribution degrees.SELECTED DRAWING: Figure 5

Description

The present invention relates to a sleep apnea syndrome determining device for determining a sleep apnea syndrome of a subject, a sleep apnea syndrome determining method, and a sleep apnea syndrome determining program.

In the medical field, the sleep state of a subject is measured in order to diagnose sleep disorders and sleep apnea syndrome. Human sleep stages are known to be classified into 6 stages from the viewpoint of sleep depth, and the 6 sleep stages are awakening, REM sleep, and non-REM sleep (stages 1 to 1) in order from the light sleep stage. It is called 4). Conventionally, these six stages of sleep are determined by attaching a large number of electrodes to the face or head of the person to be measured and measuring brain waves, eye movements, and jaw myoelectricity from the large number of electrodes. It was done by analyzing the results.

In addition, patients with sleep apnea syndrome often become apnea during sleep when the sleep stage is awake, and it is necessary to measure the sleep stage in order to diagnose apnea syndrome.
However, in order to determine sleep apnea syndrome, in addition to the measurement of the sleep stage, the measurement of the air flow accompanying breathing such as the airflow of the mouth and nose, and the ventilation exercise of the chest and abdomen, etc. It is necessary to make various measurements at the same time. Then, the doctor diagnoses whether or not he / she has apnea syndrome based on the analysis result of the sleep stage and the measurement result of the respiratory state.

The sleep test with a large number of electrodes on the face and head, which is necessary for making such a diagnosis, is usually done by staying at a medical institution and wearing the electrodes on the body continuously for a long time. This is a test performed by the patient, which imposes a mental and physical burden on the subject (patient). In addition, the acquired data needs to be analyzed and judged by a doctor with specialized knowledge and experience, and there is a problem that sleep apnea syndrome cannot be easily judged.

In order to solve the problem of measuring the sleep stage, many sleep stage estimation methods that do not require diagnosis by a specialist have been proposed.
For example, Patent Document 1 describes a technique for estimating a sleep stage from detection data of a mattress type pressure sensor by a method called Database-based Compact Genetic Algorithm, which is an improvement of a learning method using a genetic algorithm. The technique described in Patent Document 1 estimates the sleep stage based on the body movement and heartbeat of the subject detected by the mattress type pressure sensor. By estimating the sleep stage using such a mattress type pressure sensor, it is possible to estimate the sleep state of the person to be measured without imposing a burden on the person to be measured.

Japanese Unexamined Patent Publication No. 2014-239789

As mentioned above, when a patient with apnea syndrome becomes apnea during sleep, the sleep stage is often awake, so the detection data of the mattress type pressure sensor, that is, the detection data of body movement, is used. Detecting the sleep stage is one of the indicators for determining apnea syndrome.

When detecting the sleep stage from the detection data of the mattress type pressure sensor, for example, when a body movement exceeding a specific threshold is detected from the detection data of the mattress type pressure sensor, the process of determining that the sleep stage is in the awake state. It is conceivable to do. In the case of sleep of a normal person, it is possible to determine whether or not the sleep stage is awake with relatively high accuracy by performing such determination of exceeding the threshold value of body movement.

However, in the case of a patient with sleep apnea syndrome, the change pattern of body movement may be completely different from that of a normal person during sleep, and even if the same discrimination process as the sleep stage of a normal person is performed, it cannot be discriminated as awakening. Occurs a lot. Specifically, when a patient with sleep apnea syndrome is in an apnea state, even if the sleep stage is awake, the patient becomes apnea with almost no large body movement that causes the trunk to move. Therefore, there are cases in which apnea cannot be detected by determining body movement using a threshold.
Therefore, it has been difficult to directly apply the conventionally proposed process of detecting the sleep stage from the detection data of the mattress type pressure sensor for the determination of sleep apnea syndrome.

An object of the present invention is to provide an apnea syndrome determination device, an apnea syndrome determination method, and an apnea syndrome determination program capable of determining sleep apnea syndrome without imposing a burden on the subject. And.

The sleep apnea syndrome determining device of the first invention has a biometric data acquisition unit that acquires biovibration data based on the body movement or pressure change during sleep of the subject, and a biovibration data acquired by the biometric data acquisition unit. Therefore, a biometric data processing unit that calculates a plurality of feature quantities at regular intervals and a plurality of decision trees having branching conditions for determining the sleep stage from the plurality of feature quantities calculated by the biometric data processing unit are used. After discriminating the sleep stage, a determination unit for determining sleep apnea syndrome is provided based on the determination result obtained by combining the classification criteria of the decision tree used for the determination.

In addition, the method for determining sleep apnea syndrome of the first invention includes a biological data acquisition process for obtaining biological vibration data based on the body movement or pressure change during sleep of the subject, and a living body acquired by the biological data acquisition process. Using biometric data processing that calculates multiple feature quantities at regular time intervals from vibration data, and multiple decision trees that have branching conditions to determine the sleep stage from multiple feature quantities calculated by biometric data processing. After determining the sleep stage, a determination process for determining sleep apnea syndrome based on the determination result obtained by combining the classification criteria of the decision tree used for the determination is included.

In addition, the sleep apnea syndrome determination program of the first invention causes a computer to execute each process performed by the above-mentioned first sleep apnea syndrome determination method as a procedure.

The sleep apnea syndrome determining device of the second invention has a biometric data acquisition unit that acquires biovibration data based on the body movement or pressure change during sleep of the subject, and a biovibration data acquired by the biometric data acquisition unit. From, a plurality of determinations having different branching conditions for the biometric data processing unit that calculates the power spectrum of the frequency component included in the biovibration data at regular intervals and the power spectrum of the frequency component calculated by the biometric data processing unit. It is provided with a determination unit that calculates the contribution to classification by trees and determines sleep apnea syndrome based on the calculated contribution.

Further, the method for determining sleep apnea syndrome of the second invention includes a biological data acquisition process for obtaining biological vibration data based on the body movement or pressure change during sleep of the subject, and a biological data acquisition process for the living body acquired by the biological data acquisition process. Multiple determinations with different branching conditions for the biometric data processing that calculates the power spectrum of the frequency component included in the biovibration data at regular intervals from the vibration data and the power spectrum of the frequency component calculated by the biometric data processing. It includes a determination process for calculating the contribution to classification by trees and determining sleep apnea syndrome based on the calculated contribution.

In addition, the sleep apnea syndrome determination program of the second invention causes a computer to execute each process performed by the sleep apnea syndrome determination method of the second invention as a procedure.

According to the present invention, it becomes possible to accurately determine sleep apnea syndrome from biological vibration data based on body movements or pressure changes during sleep of a subject. Therefore, it is possible to easily and accurately determine sleep apnea syndrome simply by sleeping with a pressure sensor such as a mattress type.

It is a block diagram which shows the structural example of the sleep apnea syndrome determination apparatus according to the 1st Embodiment of this invention. It is a figure which shows the example of the determination state of sleep apnea syndrome by the example of 1st Embodiment of this invention. It is a block diagram which shows the hardware configuration example of the sleep apnea syndrome determination apparatus of the 1st Embodiment example of this invention. It is a figure which compared the change example of a sleep stage between a healthy person (a) and a patient (b) of sleep apnea syndrome. It is a figure which compared the characteristic example at the time of awakening (WAKE) between the healthy person (a) and the patient (b) of sleep apnea syndrome. It is a flowchart which shows the flow of the learning phase (a) and the determination phase (b) for determining the sleep apnea syndrome by the example of 1st Embodiment of this invention. It is a figure which shows the example (a) of the biological vibration data of the patient of sleep apnea syndrome, and the enlarged version (b) of a part thereof. It is a figure which shows the example of the decision tree by the 1st Embodiment example of this invention. It is a figure which shows the example of the processing state until the data which shows a feature amount is obtained from the biological vibration data by the example of 1st Embodiment of this invention. It is a figure which shows the example from the data which shows the feature amount by the 1st Embodiment example of this invention to the processing using a random forest model. It is a figure which shows the example of the count processing of the discriminant feature amount of the random forest model by the example of 1st Embodiment of this invention. It is a figure which shows the example of the determination result by the 1st Embodiment example of this invention. It is a flowchart which shows the flow of the learning phase (a) and the determination phase (b) for determining the sleep apnea syndrome determination process by the 2nd Embodiment of this invention. It is a figure which compared the contribution degree of the frequency component in the patient (a) of sleep apnea syndrome and the healthy person (b) for demonstrating the principle of the determination process by the 2nd Embodiment of this invention. .. It is a figure which shows the example from the acquisition of the biological vibration data to the construction of the random forest model by the example of the 2nd Embodiment of this invention. It is a figure which shows the example of the process of calculating the contribution degree of each frequency from the random forest model by the 2nd Embodiment example of this invention. It is a figure which shows the example which divides the frequency distribution by the 2nd Embodiment of this invention. It is a figure which shows the definition of the index by the 2nd Embodiment of this invention. It is a figure which shows the example of the determination result by the 2nd Embodiment example of this invention.

<1. Example of the first embodiment>
Hereinafter, an example of the first embodiment of the present invention will be described with reference to FIGS. 1 to 12.
[1-1. Configuration of sleep apnea syndrome determination device]
FIG. 1 is a block diagram showing the configuration of the sleep apnea syndrome determining device 10 of the present embodiment.
FIG. 2 is a diagram showing an example of a state in which the sleep apnea syndrome determination device 10 of the present embodiment is used to determine the sleep apnea syndrome.

The sleep apnea syndrome determination device 10 of the embodiment of the present embodiment acquires the biological vibration (body movement) of the person to be measured as pressure data by the mattress sensor 2. As shown in FIG. 2, for example, the mattress sensor 2 is laid on or under the mattress of the bed 1 in which the subject A sleeps, and the biological vibration of the upper body of the subject A during sleep is used as a change in pressure. To detect. The mattress sensor 2 is arranged on the mattress below the person A to be measured as an example. For example, the mattress sensor 2 may be built in the mattress.
FIG. 2 shows an example in which the sleep apnea syndrome determination device 10 is installed beside the bed 1 and the mattress sensor 2 and the sleep apnea syndrome determination device 10 are connected by a cable. For example, the mattress sensor 2 has acquired the device. The pressure data (biological vibration data) may be transmitted to the sleep apnea syndrome determination device 10 in another room by wireless transmission.
In the following description, the pressure data output by the mattress sensor 2 is referred to as biological vibration data.

As shown in FIG. 1, the sleep apnea syndrome determination device 10 includes a biological data acquisition unit 11, a biological data processing unit 12, a determination unit 13, an output unit 14, and a learning unit 15.
The biological data acquisition unit 11 performs a biological data acquisition process for acquiring biological vibration data output by the mattress sensor 2. The biological vibration data acquired by the biological data acquisition unit 11 is supplied to the biological data processing unit 12.

The biological data processing unit 12 samples the supplied biological vibration data and converts it into digital data, and calculates the average value of the digitalized biological vibration data for each predetermined time. For example, the biological data processing unit 12 acquires the average value of the biological vibration data every second. Each average value is an average of the data for the same period.

Then, the biological data processing unit 12 calculates the average value obtained at predetermined time intervals to calculate a plurality of types of feature quantities. For example, from the average value x of the biological vibration data every second, a plurality of feature quantities SD, Range, SUM, Square, LC, and RMS shown below are calculated every 30 seconds. In the following description, it is assumed that the biological data processing unit 12 obtains the feature amount every 30 seconds from the average value x of the biological vibration data every 1 second.

Each feature amount will be described in order.
The feature amount SD shows the variance of the average value x of the biological vibration data every second for 30 seconds.
The feature amount Range indicates the difference between the maximum value and the minimum value of the average value x for 30 seconds.
The feature amount SUM is an integrated value of the average value x for 30 seconds.
The feature amount Square is an integrated value of the squared value of the average value x for 30 seconds.
Feature value LC is the difference between the average values x _n of biological vibration data per second, and the average value x _n-1 of the previous one of the average value x _n (1 second ago), the integrated value of 30 seconds Is.
The feature amount RMS is obtained by averaging the integrated value of the squared value of the average value x for 30 seconds in 30 seconds to obtain the route.

These feature quantities SD, Range, SUM, Square, LC, and RMS obtained by the biological data processing unit 12 every 30 seconds are supplied to the determination unit 13 and the learning unit 15.
Here, the biological data processing unit 12 performs a labeling process for imparting a sleep stage to data indicating a plurality of feature quantities SD, Range, SUM, Square, LC, and RMS every 30 seconds, and the labeling process is performed. The data of the feature amount in which the above is performed is supplied to the determination unit 13 and the learning unit 15.
The sleep stage given here is obtained from the biological vibration data by applying the already known sleep stage estimation method. For example, the level of a specific frequency component of the biological vibration data is compared with the threshold value, and the determined sleep stage is given to the data of the feature amount for that period. Here, the process of determining the sleep stage from the biological vibration data may be a so-called sleep stage determination process for a healthy person, which does not consider the patient with sleep apnea syndrome.

Then, the learning unit 15 uses the feature amount of the biological vibration data obtained by the biological data processing unit 12, and the learning unit 15 uses a random forest, which is one of machine learning, to awaken from the feature amount data ( Generate multiple decision trees to classify whether or not it is WAKE).
The generation of the plurality of decision trees in the learning unit 15 is performed in the learning phase. The plurality of decision trees generated by the learning unit 15 are used in the determination phase, and the determination unit 13 determines the sleep stage from the determination results using the plurality of decision trees to determine whether or not it is sleep apnea syndrome. To judge.

Further, in the above-mentioned labeling process, when the biological data processing unit 12 assigns a sleep stage to a plurality of feature quantities every 30 seconds, the sleep of the subject is analyzed by analyzing data completely different from the biological vibration data. The stage may be determined and given to the feature amount data of the corresponding period. For example, the sleep stage of the person to be measured may be determined based on the data obtained from the electrodes attached to the person to be measured and the data of the sensor that detects respiration and the like.

Then, in the determination phase, the determination unit 13 uses the plurality of decision trees obtained in the learning phase, and the determination unit 13 determines whether or not it is sleep apnea syndrome from the biological vibration data obtained by the biological data processing unit 12. Determine.
At this time, the learning unit 15 extracts the rule when it is determined to be awake (WAKE) from the generated plurality of decision trees, that is, the classification route of the decision tree, and the determination unit 13 determines the number of types of the extracted rules. Determine if it is sleep apnea syndrome.

The output unit 14 outputs the result of whether or not the sleep apnea syndrome is determined by the determination unit 13. The output unit 14 is composed of, for example, a display device, and displays the determination result of sleep apnea syndrome. Alternatively, the output unit 14 may be configured as a recording device to record the determination result of sleep apnea syndrome together with the overnight sleep state and the like. Further, when the output unit 14 displays or records, not only the determination result of sleep apnea syndrome but also the determination result of the sleep stage may be displayed or recorded at the same time.

[1-2. Hardware configuration example of sleep apnea syndrome determination device]
FIG. 3 shows an example of hardware configuration when the sleep apnea syndrome determination device 10 is configured by a computer device.
The computer device C includes a CPU (Central Processing Unit) C1, a ROM (Read Only Memory) C2, and a RAM (Random Access Memory) C3 connected to the bus C8. Further, the computer device C includes a non-volatile storage C4, a network interface C5, an input device C6, and a display device C7.

The CPU C1 reads out from the ROM C2 a program code of software that realizes each function included in the biological data processing unit 12 and the determination unit 13 of the sleep apnea syndrome determination device 10 and executes the program code. Regarding the FFT process for frequency analysis of pressure data, the CPU C1 reads a program for executing the corresponding process from the ROM C2 and executes it. Variables, parameters, etc. generated during the arithmetic processing are temporarily written in the RAM C3.

As the non-volatile storage C4, for example, HDD (Hard disk drive), SSD (Solid State Drive), flexible disk, optical disk, magneto-optical disk, CD-ROM, CD-R, magnetic tape, non-volatile memory and the like are used. Be done. In this non-volatile storage C4, in addition to the OS (Operating System) and various parameters, a program for causing the computer device C to function as the sleep apnea syndrome determining device 10 is recorded. Further, the data about the sleep stage determined by the determination unit 13 is recorded in the non-volatile storage C4.

For example, a NIC (Network Interface Card) or the like is used for the network interface C5, and various data can be transmitted / received via a LAN (Local Area Network) to which terminals are connected, a dedicated line, or the like. For example, the computer device C acquires the pressure data output by the mattress sensor 2 via the network interface C5. The input device C6 is composed of a device such as a keyboard, for example, the input device C6 sets a period for determining the sleep apnea syndrome by the sleep apnea syndrome determining device 10, and indicates an instruction of a display form of the determination result. Is done. On the display device C7, the determination result of the sleep apnea syndrome is displayed by the sleep apnea syndrome determination device 10.
In addition, it is an example that the sleep apnea syndrome determination device 10 is composed of a computer device that functions as a determination device by executing a program (software), and a part or all of the processing of the sleep apnea syndrome determination device 10 is performed. You may prepare dedicated hardware to execute.

[1-3. Principle for determining sleep apnea syndrome]
Next, the principle of the sleep apnea syndrome determining device 10 for determining the sleep apnea syndrome will be described with reference to FIGS. 4 and 5.
FIG. 4 is a diagram comparing an example (a) in which a healthy person measured the sleep stage for a certain period of time and an example (b) in which a patient with sleep apnea syndrome measured the sleep stage. In the following description, a patient with sleep apnea syndrome is also referred to as a SAS patient.

In each example of FIG. 4, the horizontal axis represents the sleep time and the vertical axis represents the sleep stage. As for the sleep stages on the vertical axis, the uppermost part shows arousal (WAKE) with the lightest sleep stage, and the lower part indicates REM sleep (REM), stage 1 non-REM sleep (NREM1), and stage 2 non-REM sleep (NREM sleep). NREM2), stage 3 non-REM sleep (NREM3), and stage 4 non-REM sleep (NREM4) change in this order, indicating that a deep sleep stage is reached.
However, in the example of FIG. 4, in both the healthy subject shown in (a) and the SAS patient shown in (b), the deepest sleep stage is stage 3 non-REM sleep (NREM3), and stage 4 Non-REM sleep (NREM4) has not been reached.
As can be seen by comparing the healthy subject (a) and the SAS patient (b) in FIG. 4, the SAS patient frequently has awakening (WAKE) due to apnea.

FIG. 5 shows the difference in the biological vibration data output by the mattress sensor 2 between a healthy person and a SAS patient. FIG. 5A is an example of biological vibration data that often appears in healthy subjects when the sleep stage is awake (WAKE). FIG. 5B shows an example of biological vibration data that frequently appears in SAS patients when the sleep stage is awake (WAKE). The horizontal axis of the graph of each characteristic shown in FIG. 5 is time, and the vertical axis is the level of biological vibration.

In the case of a healthy person, as shown in FIG. 5A, when the sleep stage is awake (WAKE), a relatively high level of biological vibration is temporarily detected in conjunction with body movements such as turning over. There are many. For example, the characteristic shown on the left side of FIG. 5A is when there is a large turnover, and the characteristic shown on the right side of FIG. 5A is when there is a relatively small body movement.
On the other hand, in the case of SAS patients, when the sleep stage is awake (WAKE), biological vibrations in various changing states are detected. That is, as shown by the characteristics at the left end of FIG. 5 (b), the biological vibration may be temporarily at a high level similar to that of a healthy person, but the three characteristics on the right side of FIG. 5 (b) As shown, small movements due to apnea are detected.
In the case of SAS patients, during awakening (WAKE), the tongue blocks the airways, causing apnea, the effect of which is manifested in changes in beating (heartbeat). Then, small body movements due to apnea, which are not detected in healthy subjects, are periodically detected. However, as shown at the right end of FIG. 5B, movements equivalent to those of a healthy person may be detected during awakening (WAKE).

In the example of the present embodiment, the biological vibration at the time of awakening (WAKE) of a healthy person shown in FIG. 5 (a) and the biological vibration at the time of awakening (WAKE) of a SAS patient as shown in FIG. 5 (b). By distinguishing various biological vibrations by learning processing using a random forest, sleep apnea syndrome is determined.

[1-4. Flow of determination process for sleep apnea syndrome]
FIG. 6 is a flowchart showing a flow of a process in which the sleep apnea syndrome determining device 10 of the present embodiment determines the sleep apnea syndrome.
When determining sleep apnea syndrome, first, learning is performed in the learning phase shown in FIG. 6A, and a process of constructing a random forest model is performed. Then, in the determination phase shown in FIG. 6B, the biological vibration of the person to be measured is acquired, and a process for determining sleep apnea syndrome is performed using a random forest model.
First, the learning phase shown in FIG. 6A will be described. The biological data acquisition unit 11 starts acquiring the biological vibration data during sleep of the subject, and the biological data processing unit 12 starts acquiring the biological vibration data during sleep, and the biological data processing unit 12 starts every ta seconds for a certain period of time. The average value of the biological vibration data (pressure data) output by the mattress sensor 2 is calculated, and the calculated average value is recorded (step S11). The ta second here is, for example, 1 second.

Then, the biological data processing unit 12 calculates a plurality of feature quantities every tb second from the average value every ta seconds (step S12). The tb second is a time sufficiently longer than the ta second, for example, 30 seconds. As an example of a plurality of feature amounts here, the feature amount SD indicating the dispersion amount described above, the feature amount Range of the difference between the maximum value and the minimum value, the feature amount SUM of the integrated value, and the feature amount of the square of the integrated value are used. Square, the integrated value of the difference between the average values x _n of biological vibration data for each ta seconds and the average value x _n-1 of the previous one of the average value x LC, the integrated value of the square of the mean value for each ta seconds There is an RMS which is a value obtained by averaging the roots in tb seconds.

Next, the biological data processing unit 12 collects data on the sleep stage of the subject during the period in which the biological vibration data of the corresponding feature amount is acquired with respect to the feature amount data every tb seconds. Here, the data of the sleep stage of the person to be measured is acquired from the analysis of the biological vibration data by the biological data processing unit 12. Alternatively, data on the sleep stage of the subject may be acquired using a measuring device different from the sleep apnea syndrome determining device 10.
When the data of the sleep stage of the person to be measured is acquired in this way, the biological data processing unit 12 performs a labeling process for imparting a sleep stage to the data of the feature amount every tb seconds (step S13). The sleep stage imparted by the labeling process here is at least two stages, that is, the sleep stage is awakening (WAKE) or non-awakening (Non-WAKE).
The feature amount data obtained by the biological data processing unit 12 and labeled with the sleep stage is sent to the learning unit 15.

Then, the learning unit 15 learns using a random forest model (RF model), which is one of machine learning, and makes a plurality of decisions having different branching conditions for determining the sleep stage from a plurality of features. Build a tree (step S14). Here, when constructing a plurality of decision trees, the depth of each decision tree (the number of stages of branching conditions) and the number of decision trees to be constructed are determined in advance. For example, the depth (number of branches) of the decision tree is 10 steps, and the number of decision trees to be constructed is 300.
In addition, when determining the branching condition, the process of branching into two is roughly divided into two, for example, awakening (WAKE) and non-awakening (Non-WAKE) using an index called the Gini coefficient. It is a condition.
Further, when the learning unit 15 learns a combination of a plurality of decision trees, it determines the number of types of combinations of the plurality of feature quantities, the importance of the feature quantities, and the number of appearances of each feature quantity. Do. For example, learn which features are important for determining sleep apnea syndrome.
This is done in the learning phase, and the learning unit 15 stores the constructed random forest model (RF model).

Next, the flow of the process of determining sleep apnea syndrome from the biological vibration of the subject in the determination phase shown in FIG. 6B will be described.
In the determination phase, as in the learning phase, the acquisition process of the biological vibration of the subject every ta seconds in step S11 and the calculation process of the feature amount every tb seconds in step S12 are performed.
After that, the determination unit 13 inputs the data of the feature amount subjected to the awakening (WAKE) and the labeling process into the plurality of decision trees of the random forest model generated by the learning unit 15 in step S14 in the learning phase (step). S15).
Based on the data input to the plurality of decision trees, the determination unit 13 classifies the data of each feature amount, and the route (from the root node) used when the branch result becomes awakening (WAKE). A series of features up to the leaf node) is extracted (step S16). At this time, the determination unit 13 counts the number of determinations for each type of feature amount based on which type of feature amount data is used to determine which type of feature amount to branch at each branch of the decision tree.

Further, the determination unit 13 compares the feature amounts used in each branch of the route extracted in step S16, and determines whether or not the subject is a SAS patient based on the type and number of feature amounts (step S17). ). The feature comparison here is, for example, the number of times each of the above-mentioned plurality of feature quantities SD, Range, SUM, Square, LC, and RMS is determined at the time of branching (number of appearances), and the total of the types of feature quantities determined at the time of branching. There are a number of, etc. Regarding the number of appearances, the number of appearances of all the feature amounts may be determined, but the number of times the feature amount considered to be important among a plurality of types of feature amounts is determined is used to determine whether or not the patient is a SAS patient. It is conceivable to set a combination of classification criteria when classifying using the decision tree used for discrimination, with various conditions such as preferential treatment in.
A specific example of determining by majority voting of the number of features will be described later. Then, the output unit 14 outputs the determination result in step S17.

[1-5. Details of discrimination processing]
Next, details of each process of the flow shown in the flowchart of FIG. 6 will be described with reference to FIGS. 7 to 12.
FIG. 7 shows an example of biovibration data of a patient with sleep apnea syndrome (SAS patient). The horizontal axis of FIG. 7 is time, FIG. 7 (a) shows long-term measurement data, and FIG. 7 (b) shows a part of the long-time measurement data in an enlarged manner.

The horizontal axis of the graphs shown in FIGS. 7 (a) and 7 (b) is time, and the vertical axis shows the 6-step sleep stage change PSG and the level M of the biological vibration data. The 6-step change in sleep stage PSG was a measuring device different from the sleep apnea syndrome determination device 10 of the present embodiment, and the sleep stage of the subject was measured by a polysomnography method. It is a thing. Here, the change PSG of the six sleep stages is also referred to as the correct sleep stage.
The six sleep stages shown in FIG. 7 are, in order from the top, awakening (W), REM sleep (R), non-REM sleep stage 1 (NR1), non-REM sleep stage 2 (NR2), and non-REM sleep stage 3 ( NR3), stage 4 of non-REM sleep (NR4).

The level M of the biological vibration data shown in FIGS. 7A and 7 increases due to body movement or the like. For example, in the case of a healthy person, when the level M exceeds the threshold value TH1, the sleep stage awakens. Can be determined to be.
However, the example of FIG. 7 is the biological vibration data of the SAS patient, and as described with reference to FIG. 5, the sleep stage may be awake even when the biological vibration data is low.

The part marked with a circle (○) in the part of (a) in FIG. 7 (a) is the part where awakening (W) is detected in the correct sleep stage PSG, and the part where awakening (W) is detected is the level M of the biological vibration data. Also in comparison with the threshold value TH1, it is a place determined to be awakening (W).
On the other hand, the part marked with a cross (x) in the part of the awakening (W) in FIG. 7A is the part where the awakening (W) was detected in the correct sleep stage PSG, and the part of the biological vibration data. This is a portion that was not determined to be awakening (W) in the comparison between the level M and the threshold value TH1.
As shown in FIG. 7A, in the case of the SAS patient, various biological vibration states exist even if the sleep stage is awakening (W).
In the present embodiment, learning with a random forest model is performed so as to capture these various biological vibration states and determine whether or not the patient is a SAS patient.

FIG. 8 shows an outline of the decision tree of the random forest model.
The decision tree of the random forest model is constructed by randomly extracting the subsets S ₁ , S ₂ , ..., S _N tree from the training data set S. _SNtree is the number of decision trees to be constructed. The y _i shown at the branch point of the decision tree indicates the i-th feature quantity. That is, at the branching point, which of the plurality of feature quantities is to be determined is shown.
The plurality of decision trees are data that perform different learning for each decision tree by partially learning the data, which means that a plurality of rules are generated. That is, the waveforms of various awakening biological vibration data in the case of SAS patients are captured.

Referring to the example of FIG. 8, as for example decision tree subset S ₁ shown in the left end of FIG. 8, the first branch, the feature y ₈ is whether less than 1 is determined, the feature y ₈ 1 When it is less than (True), it is determined at the next branch whether or not the feature amount y ₇ is less than 6. Then, according to this judgment, those having a feature amount y ₇ of less than 6 are separated into those whose sleep stage is awake (W) (marked with a circle) and those whose sleep stage is other than awakening (W) (marked with a cross). ..
If the feature amount y ₈ is not less than 1 (False), it is determined in the next branch whether or not the feature amount y ₂ is less than 8. Then, in this judgment, those having a feature amount y ₂ of less than 8 are separated into those whose sleep stage is awake (W) (marked with a circle) and those whose sleep stage is other than awakening (W) (marked with a cross). ..
In this way, at the final branch of each decision tree, the sleep stage is separated into those with awakening (W) (marked with a circle) and those with a sleep stage other than awakening (marked with a circle) (marked with a cross). .. In the example of FIG. 8, two branch points are shown for the sake of simplicity, but this is an example, and one decision tree may have more branch points.

Here, when a plurality of decision trees are constructed in this way, the contribution of the feature amount used as a condition for branching at the decision tree is calculated.
The contribution Imp (y) of the feature amount is calculated by, for example, the following formula.

In the equation [Equation 1], y _i is the i-th feature quantity, N _tree is the number of decision trees, and Δ (T _j (y _i )) is the node using the feature quantity y _i in the decision tree T _j and its node. The difference in Gini coefficient after the branch of the node is shown.
As shown in FIG. 8, the Gini coefficient here is that the sleep stage is awake (○ mark) and the sleep stage is other than awakening (W) (x mark) by branching at the corresponding branch point. It is a coefficient value that is high when it is clearly separated into, and low when it is not. As the sleep stage here, the sleep stage labeled in step S13 of FIG. 6 is used.

When the difference in Gini coefficient is large, as shown in the lower left of FIG. 8, at the corresponding branch point, the sleep stage is awake (○ mark) and the sleep stage is other than awakening (W) (×). Mark) shows the state of clear separation.
On the other hand, when the difference in Gini coefficient is small, as shown in the lower left of FIG. 8, the branch at the corresponding branch point, the sleep stage is awake (○ mark) and the sleep stage is other than awake (W). (X mark) indicates the state of clear separation.

In this way, by calculating the contribution of the features required to distinguish between awake and non-awakening sleep stages based on the difference in Gini coefficient, for each feature, which By using a decision tree that branches using features, it becomes possible to determine whether or not the sleep stage is awake.

Next, the flow of determination using the determination of the contribution of the feature amount from the biological vibration data will be described with reference to FIGS. 9 to 11.
FIG. 9 shows an example of the waveform of the biological vibration data for several hours from the start of sleep (falling asleep) to waking up of one subject.
In the case of the present embodiment, a process of calculating the feature amount every 30 seconds is performed from the biological vibration data for several hours as shown in FIG. For example, the feature quantities SD, Range, SUM, Square, LC, and RMS are calculated every 30 seconds.
Then, a process of labeling the sleep stage (WAKE or Non-WALKE) at that time is performed on the data showing each feature amount every 30 seconds.

Next, as shown in FIG. 10, a process of constructing a decision tree of a random forest model (RF model) is performed from the data showing each feature amount every 30 seconds in which the sleep stage is labeled.
At this time, the depth (number of branches) of each decision tree and the total number of decision trees to be constructed are determined in advance. For example, the depth of the decision tree is 10 steps, and the total number of decision trees is 300.

Then, the feature data whose sleep stage is labeled as awakening (WAKE) is input to a plurality of decision trees of the constructed random forest model (RF model).

Next, as shown in FIG. 11, for each decision tree, the path of the decision tree passed when classifying as awakening (WAKE) is extracted, and the feature amount for which branching is determined in the middle of the path of one decision tree is determined. Count the combinations by type.
The values shown for each feature amount shown on the lower side of FIG. 11 are examples of the count values.
For example, in the example of FIG. 11, the combination of the feature amount SD and SUM is 80 times, the combination of the feature amount LC and RMS is 100 times, the combination of the feature amount SD and Range is 8 times, and the combination of the feature amount Square and SUM is 79 times. Count times, ... Here, a combination exceeding 10 times is counted as one. Then, for example, when the count number of the combination of the feature amounts of the pathway exceeds the threshold value, the patient is determined to be a SAS patient.

FIG. 12 shows an example of the determination result of whether or not the patient is a SAS patient according to the example of the present embodiment. The vertical axis of FIG. 12 shows the number of counts of the types of combinations of the feature amounts of the routes.
In the example of FIG. 12, 9 SAS patients A, B, C, ..., I and 9 healthy subjects (non-SAS patients) a, b, c, ..., I were determined. The result. Among the 18 persons shown in FIG. 12, the persons marked with x indicate the case where the determination could not be made correctly.
In FIG. 12, the threshold value of the count number of the combination of the feature amounts of the route is 100, and when the count number is 100 or more, it is determined to be a SAS patient, and when the count number is less than 100, it is not a SAS patient. It is discriminating.

In the case of the present embodiment, as can be seen from FIG. 12, 6 out of 9 SAS patients could be correctly determined to be SAS patients, and 3 were erroneously determined not to be SAS patients. On the other hand, as for healthy subjects, 7 patients were correctly determined to be healthy subjects, and 2 patients were erroneously determined to be SAS patients.
As can be seen from FIG. 12, it is possible to determine with high accuracy whether or not the patient is a SAS patient. The result shown in FIG. 12 is one simulation result for the determination, and the accuracy of the determination can be further improved by repeating the learning by the decision tree.

<2. Example of the second embodiment>
Hereinafter, examples of the second embodiment of the present invention will be described with reference to FIGS. 13 to 19.
Also in the second embodiment, the configuration described in FIGS. 1 to 3 in the first embodiment is applied as it is to the configuration of the sleep apnea syndrome determining device 10.
In the embodiment of the present embodiment, the biological data processing in the biological data processing unit 12 of the sleep apnea syndrome determination device 10, the determination processing in the determination unit 13, and the learning processing in the learning unit 15 are the first embodiments. It is different from the morphological example.

[2-1. Flow of sleep stage determination process]
FIG. 13 is a flowchart showing a flow of a process in which the sleep apnea syndrome determining device 10 of the present embodiment determines the sleep apnea syndrome.
When determining sleep apnea syndrome, first, learning is performed in the learning phase shown in FIG. 13A, and a process of constructing a random forest model is performed. Then, in the determination phase shown in FIG. 13 (b), the biological vibration of the person to be measured is acquired, and a process for determining sleep apnea syndrome is performed using a random forest model.
First, the learning phase shown in FIG. 13A will be described. The biometric data acquisition unit 11 starts acquiring the biovibration data during sleep of the subject, and the biometric data processing unit 12 starts acquiring the biovibration data during sleep, and the biometric data processing unit 12 starts every tk seconds for a certain period of time. The biological vibration data (pressure data) output by the mattress sensor 2 is acquired, and the acquired biological vibration data is recorded. The tc second here is, for example, 1 second.
Here, when calculating the biological vibration data every tk seconds (every 1 second), the biological vibration data of the past predetermined time (for example, 64 seconds) is Fourier transformed, and the power spectrum for each frequency at the predetermined time. Is calculated (step S21).

Then, the biological data processing unit 12 labels the sleep stage on the data of the power spectrum for each frequency subjected to the Fourier transform (step S22). The sleep stage imparted by the labeling process here is at least two stages, that is, the sleep stage is awakening (WAKE) or non-awakening (Non-WAKE). The labeled sleep stage is obtained from the analysis of the biological vibration data by the biological data processing unit 12, similarly to the labeled sleep stage in the first embodiment. Alternatively, data on the sleep stage of the subject may be acquired using a measuring device different from the sleep apnea syndrome determining device 10.
The power spectrum data obtained by the biological data processing unit 12 and labeled with the sleep stage is sent to the learning unit 15.

Then, the learning unit 15 learns using a random forest model (RF model), which is one of machine learning, and makes a plurality of decisions having different branching conditions for determining the sleep stage from a plurality of power spectra. Build a tree (step S23). Here, when constructing a plurality of decision trees, the depth of each decision tree (the number of stages of branching conditions) and the number of decision trees to be constructed are determined in advance.
This is done in the learning phase, and the learning unit 15 stores the constructed random forest model (RF model).

Next, the flow of the process of determining sleep apnea syndrome from the biological vibration of the subject in the determination phase shown in FIG. 13B will be described.
In the determination phase, as in the learning phase, the power spectrum calculation process for each frequency at a predetermined time in step S21 is performed.
After that, the determination unit 13 determines the contribution of each frequency component from the random forest model (RF model) generated by the learning unit 15 in step S23 in the learning phase for the data whose sleep stage is labeled as awakening (WAKE). Calculate (step S24).
After that, the determination unit 13 divides the distribution of the contribution of each frequency component into large and small by using the Gini coefficient (step S25).

Further, the determination unit 13 determines the sleep apnea syndrome from the relationship in which the distribution of the contribution of each frequency component is divided into large and small (strictly speaking, the two relationships are large and small on the x-axis and upper and lower on the y-axis). It is determined whether or not it is sleep apnea syndrome (step S26).
The output unit 14 outputs the result determined by the determination unit 13.

[2-2. Principle of discrimination from power spectrum for each frequency]
Next, with reference to FIG. 14, the difference in the contribution of the power spectrum for each frequency between the patient with sleep apnea syndrome (SAS patient) and the healthy subject will be described.
FIG. 14 (a) is data labeled as sleep apnea (WAKE), showing the contribution of the power spectrum in the case of SAS patients, and FIG. 14 (b) shows sleep apnea (WAKE). The data labeled with show the contribution of the power spectrum in the case of healthy subjects who are not SAS patients.

Here, the biological data processing unit 12 Fourier transforms the sensor value of the biological vibration data detected at 16 Hz for 64 seconds to calculate the power spectrum of [16 × 64s = 1024], and the power spectrum of the 1024 is calculated. A process of converting the 0 to 8 Hz component into a 512-dimensional vector component is performed. Here, the vector component of the 0th to 11th dimensions is 0.

The vertical axis of FIG. 14 shows the 512-dimensional vector component, and indicates the frequency position of 0 to 8 Hz. For example, the position of 100 on the vertical axis of FIG. 14 indicates a frequency position of [100 ÷ 64] = about 1.56 Hz. The horizontal axis shows the contribution of the power spectrum of each frequency component.
In the case of the example of FIG. 14, among the biological vibrations, for example, the biological vibrations associated with respiration appear mainly at positions 10 to 40, and the biological vibrations associated with heartbeat mainly appear at positions 40 to 120.

Here, in the case of the SAS patient shown in FIG. 14A, the contribution of the frequency component lower than the frequency position 100 (frequency of about 1.56 Hz) in the figure is large, and the contribution of the frequency component higher than that is large. It turns out that the degree is small.
On the other hand, in the case of the healthy person in FIG. 14 (b), the contribution of the frequency component higher than the frequency position 100 (frequency of about 1.56 Hz) in the figure is large, and the contribution of the frequency component lower than that is large. It turns out that it is small.
In the case of the present embodiment, focusing on the difference in the contribution of the power spectrum of the frequency component between the SAS patient and the healthy person shown in FIG. 14, learning using the random forest model is performed with the SAS patient and the healthy person. It is intended to identify the person.

[2-3. Details of discrimination processing]
Next, the details of the process for discriminating the SAS patient will be described with reference to FIGS. 15 to 18.
FIG. 15 shows an example of the waveform of the biological vibration data for several hours from the start of sleep (falling asleep) to waking up of one subject.
In the case of the present embodiment, the frequency of the biological vibration data (pressure data) is calculated every 64 seconds from the biological vibration data for several hours as shown in FIG. 15 to obtain a power spectrum. However, when the power spectrum is obtained in units of 64 seconds, the width for calculating the power spectrum is moved every second. The details of the process for obtaining the power spectrum of the frequency component are as described in the description of FIG.

The biological data processing unit 12 labels each of the power spectra of the frequency components as awakening (WAKE) or non-awakening (Non-WAKE).
Then, a random forest model (RF model) is constructed from the power spectra of the labeled frequency components.

As shown in FIG. 16, the biometric data processing unit 12 calculates the contribution of each frequency from the random forest model. The random forest model has a plurality of decision trees, and each decision tree has one or more branches, as described in the first embodiment. However, in the case of the second embodiment, the branching condition is the frequency.
When calculating the contribution of each frequency from the random forest model, for example, the contribution Imp ₍ y _i) is obtained from the following equation [Equation 2].

In the equation [Equation 2], y _i is the i-th frequency (numerical value on the vertical axis in FIG. 14), N _tree is the number of decision trees, and Δ (T _j (y _i )) is the frequency y in the decision tree T _j . It is the difference between the node using _i and the Gini coefficient after the branch of the node.

For example, to indicate when the difference between the Gini coefficient 16 is large, the frequency position of y _10, the branch of a decision tree, by performing the determination of y _{10 <6,} arousal indicated by ○ mark (WAKE ) Or other than awakening (Non-WAKE) indicated by x mark data can be clearly distinguished.

On the other hand, as the difference between the Gini coefficient in FIG. 16 shows a case of a small, the frequency position of y _512, with branches of a decision tree, by performing the determination of y _{512 <3,} arousal indicated by ○ mark (WAKE ) And other than the awakening (Non-WAKE) indicated by the x mark data, are indistinguishable from the decision tree.

Incidentally, Gini coefficients at each node r I _{G (r),} for example determined from the following equation [3] expression.

In the equation [Equation 3], c is the number of classes (number of classifications), and p (i | r) indicates the ratio of the elements belonging to the class i in the node r.
The difference in Gini coefficient is obtained from, for example, the following equation [Equation 4].

In the equation [Equation 4], r _L is the left side when the node r is divided, r _R is the right side when the node r is divided, and p (r _x ) belongs to the node r _x after the division. The ratio of the number of elements.

FIG. 17 shows an example in which the frequency distribution is divided into large and small (upper and lower) by using the Gini coefficient.
FIG. 17 (a) shows the case of a SAS patient, and FIG. 17 (b) shows the case of a healthy person.
In the case of the example of FIG. 17, depending on whether the contribution x on the horizontal axis is less or more than the specific position a (that is, with the contribution of x = a as a boundary), it is divided into large and small elements. Further, the frequency y on the vertical axis is divided into upper and lower (high frequency or low frequency) depending on whether the element is larger or smaller than the specific position b.

Here, in the case of the present embodiment, the process of searching for the values of the specific positions a and b in which the contribution x and the frequency y are most clearly divided by learning with a plurality of decision trees of the random forest model is performed. Do.
By determining the values of the specific positions a and b in this way, it becomes possible to distinguish between the frequency characteristics of the SAS patient described in FIG. 14 and the frequency characteristics of a healthy person, and the biological vibration data of the person to be measured will eventually be obtained. It becomes possible to determine whether or not it corresponds to.

FIG. 18 shows an example of an index SIF for determining whether or not a patient is a SAS patient.
The SAS patient determination index SIF (Spectrum Importance Feature) is defined by the following equation [Equation 5].

In the equation [Equation 5], Over _ave is the average of the upper side, Over _widht is the width of the upper side, and Under _max is the maximum value of the lower side.

FIG. 19 shows an example of the determination result of whether or not the patient is a SAS patient according to the example of the present embodiment. The vertical axis of FIG. 19 shows the value of the index SIF.
Here, the threshold value of the index SIF is 6000, and when the threshold value is 6000 or more, it is determined to be a SAS patient, and when the threshold value is less than 6000, it is determined not to be a SAS patient.
In the example of FIG. 19, 9 SAS patients A, B, C, ..., I and 9 healthy subjects (non-SAS patients) a, b, c, ..., I were determined. The result is shown. In the case of FIG. 19, SAS patients A to I and healthy subjects a to i are shown in numerical order of the index SIF.

In the example shown in FIG. 19, all nine SAS patients A to I are identified as SAS patients, and nine healthy subjects a to i are identified as non-SAS patients.
As can be seen from FIG. 19, according to the example of the present embodiment, it is possible to discriminate a patient with sleep apnea syndrome with extremely high accuracy.

<3. Modification example>
It should be noted that the feature amounts and frequency values described in the above-described embodiments are shown as suitable examples, and are not limited to these feature values and frequency values.

1 ... Bed, 2 ... Mattress sensor, 10 ... Sleep aspiration syndrome determination device, 11 ... Biological data acquisition unit, 12 ... Biological data processing unit, 13 ... Judgment unit, 14 ... Output unit, 15 ... Learning unit, A ... Subject, C ... computer device, C1 ... CPU, C2 ... ROM, C3 ... RAM, C4 ... non-volatile storage, C5 ... network interface display, C6 ... input device, C7 ... display device, C8 ... bus

Claims (8)

A biometric data acquisition unit that obtains biovibration data based on body movements or pressure changes during sleep of the subject,
A biological data processing unit that calculates a plurality of feature quantities at regular intervals from the biological vibration data acquired by the biological data acquisition unit.
After discriminating the sleep stage using a plurality of decision trees having branching conditions for discriminating the sleep stage from the plurality of feature quantities calculated by the biometric data processing unit, the classification of the decision tree used for the discrimination. A sleep apnea syndrome determination device including a determination unit for determining sleep apnea syndrome based on the determination result obtained by a combination of criteria. It is equipped with a learning unit that learns the multiple decision trees.
The method according to claim 1, wherein in the learning of a combination of a plurality of decision trees by the learning unit, information on the sleep stage that is the correct answer is acquired, and a decision tree that can determine the sleep stage that has a high degree of agreement with the correct answer is selected. Sleep apnea syndrome determination device. In claim 2, when the determination unit determines sleep apnea syndrome, the number of types of combination of feature amounts and the number of appearances of each feature amount when branching by the plurality of decision trees are used. The sleep apnea syndrome determination device described. A biometric data acquisition unit that obtains biovibration data based on body movements or pressure changes during sleep of the subject,
A biometric data processing unit that calculates the power spectrum of a frequency component included in the biovibration data at regular intervals from the biovibration data acquired by the biometric data acquisition unit.
Regarding the power spectrum of the frequency component calculated by the biological data processing unit, the contribution to the classification by a plurality of decision trees having different branching conditions is calculated, and the sleep apnea syndrome is determined based on the calculated contribution. A sleep apnea syndrome determination device equipped with a determination unit for performing. Biometric data acquisition processing to obtain biovibration data based on body movement or pressure change during sleep of the subject,
Biometric data processing that calculates a plurality of feature quantities at regular intervals from the biovibration data acquired by the biometric data acquisition process.
After discriminating the sleep stage using a plurality of decision trees having branching conditions for discriminating the sleep stage from the plurality of feature quantities calculated by the biometric data processing, the classification criteria of the decision tree used for the judgment. A method for determining sleep apnea syndrome, which includes a determination process for determining sleep apnea syndrome based on the discrimination results obtained by the combination of. Biometric data acquisition processing to obtain biovibration data based on body movement or pressure change during sleep of the subject,
Biometric data processing that calculates the power spectrum of the frequency component included in the biovibration data at regular intervals from the biovibration data acquired by the biometric data acquisition process.
Regarding the power spectrum of the frequency component calculated by the biometric data processing, the contribution to the classification by a plurality of decision trees having different branching conditions is calculated, and the sleep apnea syndrome is determined based on the calculated contribution. Judgment processing to be performed and sleep apnea syndrome determination method including. Biometric data acquisition procedure to obtain biovibration data based on body movement or pressure change during sleep of the subject,
A biometric data processing procedure that calculates a plurality of feature quantities at regular intervals from the biovibration data acquired by the biometric data acquisition procedure, and
After discriminating the sleep stage using a plurality of decision trees having branching conditions for discriminating the sleep stage from the plurality of feature quantities calculated in the biometric data processing procedure, the classification of the decision tree used for the judgment is performed. Judgment procedure for determining sleep apnea syndrome based on the discrimination results obtained by combining the criteria, and
A sleep apnea syndrome determination program that lets a computer run. Biometric data acquisition procedure to obtain biovibration data based on body movement or pressure change during sleep of the subject,
A biological data processing procedure for calculating the power spectrum of a frequency component included in the biological vibration data at regular intervals from the biological vibration data acquired by the biological data acquisition procedure, and a biological data processing procedure.
For the power spectrum of the frequency component calculated by the biological data processing procedure, the contribution to the classification by a plurality of decision trees having different branching conditions is calculated, and the sleep apnea syndrome is determined based on the calculated contribution. Judgment procedure to perform and
A sleep apnea syndrome determination program that lets a computer run.