JP7614584B2 - DETECTION APPARATUS, DETECTION METHOD, AND PROGRAM - Google Patents

Description

The present invention relates to a detection device, a detection method, and a program.

Conventionally, it has been considered to detect symptoms of diseases such as heart failure based on electrocardiogram information and pulse wave information. For example, Patent Document 1 describes a method for detecting the occurrence of atrial fibrillation based on distribution information that indicates the magnitude of fluctuation in pulse intervals. Using a method such as that described in Patent Document 1, it is possible to detect abnormal heartbeats, which are symptoms of diseases such as heart failure, early and provide effective treatment, without waiting for the patient to become aware of the symptoms and voluntarily visit a doctor.

JP 2016-202346 A

In such detection methods, there is a demand for the ability to properly detect signs of recurrence of diseases such as heart failure. The present invention has been made to solve the above-mentioned problems, and aims to provide a detection device, detection method, and program that properly detects abnormal heartbeats that are signs of recurrence of heart failure.

The detection device according to the present invention is characterized by having an acquisition means for acquiring pulse wave information of a patient who has been cured of an illness and the measurement date and time of the pulse wave information, a setting means for acquiring the date and time of cure when the illness was cured, setting a stable period near the date and time of cure, and setting a target period after the stable period, a first calculation means for calculating a first characteristic amount based on pulse wave information whose measurement date and time falls within the stable period, a second calculation means for calculating a second characteristic amount based on pulse wave information whose measurement date and time falls within the target period, a detection means for detecting an abnormality in the patient based on the distance between the first characteristic amount and the second characteristic amount, and an output means for outputting the detection result.

In addition, in the detection device according to the present invention, it is preferable that the setting means sets a plurality of stable periods, the first calculation means calculates a first feature amount for each of the plurality of stable periods, and the detection means detects an abnormality in the patient based on the distance between a feature amount group consisting of the plurality of first feature amounts and the second feature amount.

In addition, in the detection device according to the present invention, it is preferable that the setting means sets a plurality of target periods, the second calculation means calculates the second feature amount for each of the plurality of target periods, and the detection means detects an abnormality in the patient based on the number of second feature amounts that have a distance between the second feature amount and the first feature amount that is equal to or greater than a predetermined value.

The detection device according to the present invention further includes a motion state acquisition means for acquiring the motion state of the patient during the stable period and a plurality of target periods, and an extraction means for extracting a target period in which a motion state similar to the stable period was acquired from among the plurality of target periods, and the detection means preferably detects an abnormality in the patient based on the number of second feature amounts, among the second feature amounts calculated for the extracted target periods, whose distance from the first feature amount is equal to or greater than a predetermined value.

The detection device according to the present invention preferably further comprises a sleep state acquisition means for acquiring a sleep state indicating whether the patient is asleep or not, and the setting means sets the stable period and the target period so that they are included in the time when the patient is asleep.

In addition, in the detection device according to the present invention, it is preferable that the acquisition means acquires the patient's electrocardiogram waveform as pulse wave information, the setting means acquires the patient's discharge date and time as the recovery date and time, and the first calculation means and the second calculation means calculate, as the first feature amount and the second feature amount, parameters of the distribution of two-dimensional data obtained by Lorentz plotting the RR interval of the electrocardiogram waveform.

The detection method according to the present invention is a detection method executed by a computer, and is characterized in that it includes acquiring pulse wave information of a patient who has been cured of a disease and the measurement date and time of the pulse wave information, acquiring the cure date and time when the disease was cured, setting a stable period near the cure date and time and setting a target period after the stable period, calculating a first feature amount based on the pulse wave information whose measurement date and time fall within the stable period, calculating a second feature amount based on the pulse wave information whose measurement date and time fall within the target period, and detecting an abnormality in the patient based on the distance between the first feature amount and the second feature amount.

The program according to the present invention causes a computer to execute the following operations: acquire pulse wave information of a patient who has been cured of an illness and the measurement date and time of the pulse wave information; acquire the cure date and time when the illness was cured; set a stable period near the cure date and time and set a target period after the stable period; calculate a first feature amount based on the pulse wave information whose measurement date and time falls within the stable period; calculate a second feature amount based on the pulse wave information whose measurement date and time falls within the target period; and detect an abnormality in the patient based on the distance between the first feature amount and the second feature amount.

The detection device, detection method, and program of the present invention make it possible to properly detect abnormal heartbeats that are a sign of recurrence of heart failure.

FIG. 1 is a diagram illustrating an example of a schematic configuration of a detection system 1. FIG. 2 is a diagram showing an example of a schematic configuration of a detection device 2. 4 is a diagram showing an example of the data structure of an electrocardiogram waveform table 211. FIG. 4 is a diagram showing an example of the data structure of an exercise state table 212. FIG. 4 is a diagram showing an example of the data structure of an RR interval table 213. FIG. FIG. 4 is a diagram showing an example of the data structure of a feature amount table 214. FIG. 11 is a flow diagram showing an example of the flow of a detection process. FIG. 11 is a schematic diagram for explaining calculation of a feature amount.

Below, embodiments of the present invention will be described with reference to the drawings. Please note that the technical scope of the present invention is not limited to these embodiments, but extends to the inventions described in the claims and their equivalents.

Fig. 1 is a diagram showing an example of a schematic configuration of a detection system 1 according to the present invention. The detection system 1 is used, for example, to detect abnormal heartbeats that are a sign of recurrence of a circulatory system disease, such as heart failure, in a patient who has been cured of the disease. The detection system 1 has a detection device 2, an electrocardiograph 3, and an activity meter 4. The detection device 2, the electrocardiograph 3, and the activity meter 4 are connected to each other via a network 5 so as to be able to communicate with each other.

The electrocardiograph 3 is attached to the patient's chest or the like and measures the patient's electrocardiogram waveform. The electrocardiograph 3 is preferably a Holter-type electrocardiograph capable of measuring for several hours or more, but may be any other electrocardiograph. The electrocardiograph 3 has a communication function and transmits time-series data indicating the patient's electrocardiogram waveform to the detection device 2 via the network 5. The electrocardiogram waveform is an example of pulse wave information.

The activity meter 4 is attached to the patient's limbs, etc., and measures the patient's motion and sleep states. The motion state is data indicating the magnitude of the patient's body movement at each measurement time, and the sleep state is data indicating whether the patient is asleep at each measurement time. The activity meter 4 measures the patient's motion and sleep states as described below. The activity meter 4 has a communication function, and transmits data indicating the patient's motion and sleep states to the detection device 2 via the network 5.

FIG. 2 is a diagram showing an example of a schematic configuration of the detection device 2. The detection device 2 is an information processing device such as a PC (Personal Computer), a server, a mobile phone, a multi-function mobile phone (smartphone), a tablet terminal, etc. The detection device 2 calculates features based on the electrocardiogram waveform measured by the electrocardiograph 3, and detects abnormalities in the patient based on the calculated features. To this end, the detection device 2 has a memory unit 21, a communication unit 22, and a processing unit 23.

The storage unit 21 is configured to store data and programs, and includes, for example, a semiconductor memory. The storage unit 21 stores operating system programs, driver programs, application programs, data, etc., used for processing by the processing unit 23. Programs are installed into the storage unit 21 using a setup program or the like from a computer-readable, non-transient portable storage medium such as a CD (Compact Disc)-ROM (Read Only Memory).

The memory unit 21 stores an electrocardiogram waveform table 211, a motion state table 212, an RR interval table 213, and a feature table 214 as data.

The communication unit 22 is configured to enable the detection device 2 to communicate with other devices via a network such as the Internet or an intranet, and includes, for example, a communication interface circuit. The communication interface circuit included in the communication unit 22 is, for example, a communication interface circuit for a wired LAN (Local Area Network) or a wireless LAN. The communication unit 22 transmits data supplied from the processing unit 23 to other devices, and also supplies data received from other devices to the processing unit 23.

The processing unit 23 is configured to comprehensively control the operation of the detection device 2, and includes one or more processors and their peripheral circuits. The processing unit 23 includes, for example, a CPU (Central Processing Unit), an LSI (Large Scale Integration), or an ASIC (Application Specific Integrated Circuit). The processing unit 23 may also include a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), etc. The processing unit 23 controls the operation of each component of the detection device 2 so that the processing of the detection device 2 is performed appropriately based on the program stored in the memory unit 21, and executes the processing.

The processing unit 23 has a first acquisition means 231, a second acquisition means 232, a setting means 233, an extraction means 234, a first calculation means 235, a second calculation means 236, a detection means 237, and an output means 238 as functional blocks. These functional blocks are functional modules realized by a program executed by the processing unit 23. These functional blocks may be realized by being implemented in the detection device 2 as firmware. The second acquisition means is an example of an exercise state acquisition means and a sleep state acquisition means.

Figure 3 is a diagram showing an example of the data structure of the electrocardiogram waveform table 211 stored in the memory unit 21. The electrocardiogram waveform table 211 stores electrocardiogram waveforms and their measurement dates and times in association with each other. The electrocardiogram waveform is time-series data indicating cardiac activity. In the example shown in Figure 3, the electrocardiogram waveform is data indicating the change over time in the electric potential generated at the electrodes of the electrocardiograph 3 attached to the patient's body in association with cardiac activity.

The electrocardiograph 3 generates data for the electrocardiogram waveform table 211 by measuring the patient's electrocardiogram waveform over a predetermined time period (e.g., from 10 p.m. to 5 a.m. the following morning) for each measurement day. The detection device 2 receives the data for the electrocardiogram waveform table 211 in advance from the electrocardiograph 3 via the communication unit 22 and stores it in the memory unit 21.

Figure 4 is a diagram showing an example of the data structure of the exercise state table 212 stored in the memory unit 21. The exercise state table 212 stores the exercise state, sleep state, and their measurement date and time in association with each other. The exercise state is time series data indicating the magnitude of the patient's body movement at each measurement time. The sleep state is time series data indicating whether the patient is sleeping or not at each measurement time.

The activity meter 4 generates data for the exercise state table 212 by measuring the patient's exercise state and sleep state over a predetermined time period for each measurement day. The activity meter 4 is equipped with an acceleration sensor, and measures the patient's exercise state and sleep state by acquiring the acceleration of the patient's limbs on which the activity meter 4 is worn.

For example, if the acceleration sensor detects acceleration equal to or greater than the first threshold value before and after the measurement time, the activity meter 4 generates data indicating that the patient was not asleep and had large body movements at that measurement time. If the acceleration sensor detects acceleration less than the first threshold value and equal to or greater than the second threshold value before and after the measurement time, the activity meter 4 generates data indicating that the patient was not asleep and had small body movements at that measurement time. If the acceleration sensor detects only acceleration less than the second threshold value before and after the measurement time, the activity meter 4 generates data indicating that the patient was asleep at that measurement time.

The activity meter 4 may include sensors for measuring body temperature and pulse in addition to the acceleration sensor, and generate data on the patient's exercise state and sleep state based on the patient's body temperature and pulse. The activity meter 4 may also generate, as the exercise state, numerical data corresponding to the magnitude of the detected acceleration, and, as the sleep state, numerical data corresponding to the estimated depth of sleep.

Figure 5 is a diagram showing an example of the data structure of the RR interval table 213 stored in the storage unit 21. The RR interval table 213 stores RR intervals and their measurement dates and times in association with each other. The RR interval is the time interval between adjacent R waves in an electrocardiogram waveform. In the example shown in Figure 5, the measurement time is divided into five-minute intervals, and the average value of the RR interval over the five minutes is stored. The data in the RR interval table 213 is generated by the detection device 2 based on the electrocardiogram waveform table 211 in the detection process described below.

Figure 6 is a diagram showing an example of the data structure of the feature amount table 214 stored in the storage unit 21. The feature amount table 214 stores periods and feature amounts in association with each other. The data of the feature amount table 214 is generated by the detection device 2 based on the RR interval table 213 in the detection process described below. In the example shown in Figure 6, the feature amount is calculated for each measurement date. The content of the feature amount and the calculation method will be described later.

Figure 7 is a flow diagram showing an example of the flow of the detection process executed by the detection device 2. The detection process is executed when detecting an abnormality in a patient, with the above-mentioned tables and the date and time of the patient's recovery stored in advance in the memory unit 21. The detection process is realized by the processing unit 23 working in cooperation with each component of the detection device 2 based on a program executed by the memory unit 21.

First, the first acquisition means 231 acquires the electrocardiogram waveform of a patient whose disease has been cured and the measurement date and time of the electrocardiogram waveform (S11). The first acquisition means 231 acquires the electrocardiogram waveform table 211 from the storage unit 21 as the patient's electrocardiogram waveform and its measurement date and time.

Next, the second acquisition means 232 acquires the patient's exercise state and sleep state (S12). The second acquisition means 232 acquires the exercise state table 212 from the memory unit 21 as the patient's exercise state and sleep state.

Next, the setting means 233 acquires the date and time of cure when the patient's disease was cured, sets a stable period near the date and time of cure, and sets a target period after the stable period (S13).

The setting means 233 acquires the patient's discharge date and time from the memory unit 21 as the recovery date and time. The setting means 233 sets a stable period immediately after the discharge date and time. For example, the setting means 233 sets the period during which the patient is asleep as the stable period for each day of a specified period (e.g., 14 days) starting from the day after the discharge date and time, based on the sleep state stored in the exercise state table 212.

For example, if the patient is asleep for the entirety of a specified time on a certain measurement date (e.g., from 10 p.m. to 5 a.m. the following morning), the setting means 233 sets the specified time as the stable period for that measurement date. If the patient is asleep for a portion of the specified time (e.g., from midnight to 5 a.m. the following morning), the setting means 233 sets the portion of the time as the stable period for that measurement date. If the patient is not asleep for the entirety of the specified time, the setting means 233 does not set a stable period for that measurement date.

The stable period is not limited to the period starting the day after the discharge date and time as described above, so long as it is a period during which there is expected to be no significant change in the patient's condition. For example, the setting means 233 may set the stable period to a period starting on a date separated from the date and time of cure, such as three days after the date and time of cure. In addition, multiple stable periods may be set on days separated from each other. For example, the setting means 233 may set the stable period to every other day.

The date and time of recovery is not limited to the date and time of discharge as described above, but may be the date and time when a doctor diagnoses that the patient has been cured of heart failure. In this case, the setting means 233 may set the period including the period immediately prior to the date and time of recovery (e.g., the day before the date and time of recovery) as the stable period.

The setting means 233 sets a target period after the set stable period. For example, for each day of a predetermined period (e.g., 14 days) starting from the day after the last day of the stable period, the setting means 233 sets the period during which the patient is asleep as the target period based on the sleep state stored in the exercise state table 212. It is preferable that the target period is set to be included in the same time period as the stable period (e.g., from 10 p.m. to 5 a.m. the following morning).

The setting means 233 may set the target period to a period starting on a date separated from the stable period (for example, one month after the last day of the stable period). In addition, multiple target periods may be set on dates separated from each other.

Next, the extraction means 234 refers to the motion state table 212 and extracts, from among the multiple target periods, a target period in which a motion state similar to the stable period was obtained (S14). In the example shown in FIG. 4, either information indicating that the patient's body movement is large or information indicating that the patient's body movement is small is stored as the motion state. In this case, if the difference between the proportion of time during which the patient's body movement is indicated as large in a certain target period and the proportion of time during which the patient's body movement is indicated as large in a stable period is equal to or less than a predetermined value, the extraction means 234 extracts the target period as a target period in which a motion state similar to the stable period was obtained.

Then, the first calculation means 235 calculates a first feature based on the electrocardiogram waveform whose measurement date and time is included in the stable period (S15). For each stable period, the first calculation means 235 acquires the electrocardiogram waveform whose measurement date and time is included in each stable period from the electrocardiogram waveform table 211. The first calculation means 235 identifies the occurrence time of the R wave in the acquired electrocardiogram waveform and calculates the RR interval. The first calculation means 235 divides each stable period into unit periods of a predetermined time (e.g., 5 minutes), and calculates the average value of the RR intervals of the R waves occurring in each unit period as the RR interval of that unit period. The first calculation means 235 stores the calculated RR intervals of each unit period in the storage unit 21 as the RR interval table 213.

The first calculation means 235 calculates the parameters of the distribution of the two-dimensional data obtained by Lorentz plotting the calculated RR intervals as the first feature.

Figure 7 is a schematic diagram for explaining the calculation of the feature quantity. The average value of the RR interval in the nth unit period of each stable period is R(n), and the average value of the RR interval in the n+1th unit period is R(n+1). The first calculation means 235 plots a point with coordinates (R(n), R(n+1)) for each unit period on the XY plane. This provides a two-dimensional data distribution for each stable period that is located approximately along a straight line L expressed by the equation y = x, as shown in Figure 7.

The first calculation means 235 calculates the average of the obtained two-dimensional data, the variance σ+ in the direction along the line L, and the variance σ- in the direction perpendicular to the line L. The variances σ+ and σ- correspond to the variances of the x and y coordinates, respectively, when the two-dimensional data is rotated -45 degrees around the origin. The first calculation means 235 calculates the average value of the two-dimensional data and the value of π×(σ+)×(σ-), which are parameters of the distribution of the two-dimensional data, as the first feature amount in each stable period. The average value corresponds to the center of the ellipse O when the shape of the distribution of the obtained two-dimensional data is regarded as an ellipse O, and π, σ+, and σ- correspond to the area of the ellipse O, so hereinafter these feature amounts may be referred to as the center and area, respectively.

The first calculation means 235 stores the first feature calculated for each stable period in the feature table 214 in association with the measurement date of each stable period. As shown in FIG. 7, since the center of the ellipse is approximately located on a straight line expressed by the formula y=x, the first calculation means 235 may store only either the x coordinate or the y coordinate as the center of the ellipse.

The second calculation means 236, in the same manner as the first calculation means 235, calculates the second feature based on the electrocardiogram waveform whose measurement date and time are included in each extracted target period (S16). The second calculation means 236 stores the calculated second feature in the feature table 214 in association with the measurement date of each target period.

Then, the detection means 237 detects an abnormality in the patient based on the distance between the first feature amount and the second feature amount (S17). For example, the detection means 237 calculates the distance between a feature amount group consisting of a plurality of first feature amounts calculated for each of a plurality of stable periods and a second feature amount calculated for each target period extracted by the extraction means 234. The detection means 237 determines whether each calculated distance is equal to or greater than a threshold value determined based on a predetermined detection level.

The distance between the feature group and the second feature is, for example, the Mahalanobis distance d(x) expressed by the following formula.

where x is a column vector indicating the second feature, and in the above example, is a two-variable column vector whose elements are the centers and areas of the ellipses calculated for the specified target period. μ is a column vector indicating the average of the first feature, and in the above example, is a two-variable column vector whose elements are the average values of the centers and areas of the ellipses calculated for each of the multiple stable periods. Σ is a variance-covariance matrix of the first feature, and in the above example, is a 2-row, 2-column matrix whose elements are the variances of the centers and areas of the ellipses and the covariance between the centers and areas calculated for each of the multiple stable periods. The superscript T indicates transpose, and the superscript -1 indicates an inverse matrix.

The threshold is calculated as follows. That is, if it is assumed that the feature quantity follows a multidimensional normal distribution (two-dimensional normal distribution in the above example), the distribution of the square of the Mahalanobis distance is approximated by a chi-square distribution. Therefore, the threshold t for detecting anomalies with a preset detection level α (for example, 0.95 when detecting 5 percent anomalies) is calculated based on the following formula. Here, χ ² _M (u) is the probability density function of the chi-square distribution with M degrees of freedom.

The detection means 237 detects abnormalities in the patient based on the number of second feature amounts for which the distance is equal to or greater than the threshold value among the second feature amounts for each target period. For example, the detection means 237 detects abnormalities in the patient when the number of second feature amounts for which the distance is equal to or greater than the threshold value is equal to or greater than a predetermined number. The predetermined number is determined according to the number of target periods (i.e., the total number of second feature amounts), such as two per seven-day target period. The predetermined number may be determined regardless of the number of target periods.

Next, the output means 238 outputs the result of the detection by the detection means 237 (S18) and ends the detection process. For example, the output means 238 outputs the result of the detection by transmitting information indicating whether or not an abnormality has been detected to another device via the communication unit 22.

As described above, the detection device 2 sets a stable period near the date and time when the patient's disease was cured, and sets a target period after the stable period. Furthermore, the detection device 2 detects abnormalities in the patient based on the distance between a first feature based on pulse wave information whose measurement date and time falls within the stable period, and a second feature based on pulse wave information whose measurement date and time falls within the target period. This enables the detection device 2 to appropriately detect heart rate abnormalities that are a sign of recurrence of heart failure.

That is, normal heart rates generally differ from patient to patient. Detection device 2 detects abnormal heart rates by comparing each patient's heart rate with the patient's heart rate during a stable period near the time of recovery, when there is a high probability that heart failure has not recurred. This allows detection device 2 to appropriately detect abnormal heart rates even for multiple patients with different normal heart rates.

The detection device 2 also sets multiple stable periods and detects abnormalities in the patient based on the distance between a feature group consisting of multiple first feature amounts calculated for each of the multiple stable periods and the second feature amount. This enables the detection device 2 to more appropriately detect heart rate abnormalities by taking into account the distribution of feature amounts during the stable periods.

The detection device 2 also sets multiple target periods and detects abnormalities in the patient based on the number of second feature quantities that have a distance between them and the first feature quantities that is equal to or greater than a predetermined value. This enables the detection device 2 to appropriately detect heart rate abnormalities while reducing the possibility of falsely detecting heart rate abnormalities.

The detection device 2 also extracts, from among the multiple target periods, a target period in which a motion state similar to that of the stable period was acquired, and detects abnormalities in the patient based on the second feature calculated for the extracted target period. This enables the detection device 2 to more appropriately detect heart rate abnormalities using only pulse wave information measured under conditions similar to those of the stable period.

The detection device 2 also sets the stable period and the target period to be included in the time when the patient is sleeping. This enables the detection device 2 to more appropriately detect heart rate abnormalities using only pulse wave information that is unlikely to be affected by the patient's movements.

Detection device 2 also sets a stable period immediately after the patient's discharge date and time. When a patient is hospitalized, their lifestyle changes significantly during hospitalization compared to after they are discharged. Therefore, if pulse wave information measured during hospitalization is used, there is a possibility that abnormalities will be erroneously detected due to fluctuations in heart rate associated with changes in lifestyle. By setting a stable period immediately after the discharge date and time, detection device 2 makes it possible to properly detect abnormal heart rate without being affected by changes in lifestyle.

Furthermore, the detection device 2 calculates parameters of the distribution of two-dimensional data obtained by Lorentz plotting the RR interval as the first feature amount and the second feature amount. This enables the detection device 2 to appropriately detect heartbeat abnormalities based on features related to the time change of the RR interval. In particular, the center of the ellipse used by the detection device 2 as a feature amount in the above example represents the change in the RR interval on a daily measurement basis, and σ- and σ+ used to calculate the area represent the variation in the RR interval on a five-minute basis. Therefore, by using the center and area of the ellipse, the detection device 2 can appropriately detect heartbeat abnormalities based on changes in the heartbeat in various time units.

The detection device 2 may not acquire the motion state in S12 of the detection process. In this case, the detection device 2 does not execute S14. The detection device 2 may not acquire the sleep state in S12. In this case, the detection device 2 may set the stable period and the target period to a predetermined time regardless of whether the patient is asleep or not in S13.

In the above description, the first calculation means 235 and the second calculation means 236 calculate the center and area of an ellipse obtained by Lorentz plotting the RR interval as the feature amount, but this is not limited to this example. For example, the first calculation means 235 and the second calculation means 236 may use the average value and standard deviation (or variance value) of the RR interval as the feature amount. Also, the first calculation means 235 and the second calculation means 236 may use the median and interquartile range of the RR interval as the feature amount. Also, the most frequent value may be used instead of the median, and the range (the difference between the maximum value and the minimum value) may be used instead of the interquartile range. All of the above-mentioned examples of feature amounts are combinations of a representative value of the RR interval and an index of variation, but an index such as kurtosis or skewness of the distribution of the RR interval may be further used as a feature amount.

In the above description, the first calculation means 235 and the second calculation means 236 calculate the feature amount using the 5-minute average of the RR interval, but this is not limited to an example. The first calculation means 235 and the second calculation means 236 may use the average of the RR interval over an arbitrary period. The first calculation means 235 and the second calculation means 236 may use the measured RR interval as is without taking the average.

In the above description, the detection means 237 detects an abnormality based on the Mahalanobis distance, but this is not limited to the example. The detection means 237 may detect an abnormality based on the Euclidean distance or Manhattan distance between the first feature amount and the second feature amount. For example, the detection means 237 may detect an abnormality based on the Euclidean distance or Manhattan distance using the k-nearest neighbor method, the local outlier method, or the OCSVM (One-Class Support Vector Machine), etc. Furthermore, the detection means 237 may detect an abnormality based on whether the Euclidean distance between the first feature amount and the second feature amount is equal to or greater than a threshold value.

In the above description, the setting means 233 sets multiple stable periods, but this is not limited to the example, and the setting means 233 may set only one stable period. Also, the setting means 233 may set only one target period.

In the above description, the first calculation means 235 and the second calculation means 236 calculate the feature amount based on the electrocardiogram waveform, but this is not limited to the example. The first calculation means 235 and the second calculation means 236 may calculate the feature amount based on a pulse wave signal measured using a photoplethysmograph or the like.

Those skilled in the art should understand that various changes, substitutions, and modifications can be made to the present invention without departing from the spirit and scope of the present invention. For example, the processing of each part described above may be performed in a different order as appropriate within the scope of the present invention. Furthermore, the above-described embodiments and modifications may be implemented in combination as appropriate within the scope of the present invention.

2 Detection device 231 First acquisition means 232 Second acquisition means 233 Setting means 234 Extraction means 235 First calculation means 236 Second calculation means 237 Detection means 238 Output means

Claims (7)

An acquisition means for acquiring pulse wave information of a patient whose disease has been cured and the measurement date and time of the pulse wave information;
A setting means for acquiring a date and time when the disease was cured, setting a plurality of stable periods in the vicinity of the date and time when the disease was cured, and setting a target period after the stable periods;
a first calculation means for calculating a first characteristic amount based on the pulse wave information whose measurement date and time is included in the stable period;
a second calculation means for calculating a second characteristic amount based on the pulse wave information whose measurement date and time is included in the target period;
a detection means for detecting an abnormality in the patient based on a Mahalanobis distance between a distribution of a group of feature amounts consisting of the first feature amounts for each of the plurality of stable periods and the second feature amount;
an output means for outputting the result of the detection;
A detection device comprising: The setting means sets a plurality of the target periods,
The second calculation means calculates the second feature amount for each of the plurality of target periods;
the detection means detects an abnormality in the patient based on the number of second feature amounts whose Mahalanobis distance from a distribution of the group of feature amounts is equal to or greater than a predetermined value, among the second feature amounts.
The detection device according to claim 1 . an exercise status acquisition means for acquiring an exercise status of the patient during the stable period and the plurality of target periods;
and an extraction means for extracting a target period in which a motion state similar to the stable period has been acquired from the plurality of target periods,
the detection means detects an abnormality in the patient based on the number of second feature amounts, among the second feature amounts calculated for the extracted target period, whose Mahalanobis distance from a distribution of the group of feature amounts is equal to or greater than a predetermined value.
3. The detection device according to claim 2 . The sleep state acquiring means further acquires a sleep state indicating whether the patient is asleep or not,
the setting means sets the stable period and the target period so as to be included in a time during which the patient is sleeping.
A detection device according to any one of claims 1 to 3 . The acquiring means acquires an electrocardiogram waveform of the patient as the pulse wave information,
The setting means acquires a discharge date and time of the patient as the recovery date and time,
the first calculation means and the second calculation means calculate, as the first feature amount and the second feature amount, parameters of a distribution of two-dimensional data obtained by performing a Lorentz plot of an R-R interval of the electrocardiogram waveform.
A detection device according to any one of claims 1 to 4 . 1. A computer implemented detection method comprising:
Obtaining pulse wave information of a patient whose disease has been cured and the measurement date and time of the pulse wave information;
Obtaining a cure date and time when the disease is cured, setting a plurality of stable periods in the vicinity of the cure date and time, and setting a target period after the stable periods;
calculating a first characteristic amount based on the pulse wave information whose measurement date and time is included in the stable period;
calculating a second characteristic amount based on the pulse wave information whose measurement date and time is included in the target period;
detecting an abnormality in the patient based on a Mahalanobis distance between a distribution of a feature group consisting of the first feature amounts for each of the plurality of stable periods and the second feature amount;
outputting the result of said detection;
A detection method comprising: Obtaining pulse wave information of a patient whose disease has been cured and the measurement date and time of the pulse wave information;
Obtaining a cure date and time when the disease is cured, setting a plurality of stable periods in the vicinity of the cure date and time, and setting a target period after the stable periods;
calculating a first characteristic amount based on the pulse wave information whose measurement date and time is included in the stable period;
calculating a second characteristic amount based on the pulse wave information whose measurement date and time is included in the target period;
detecting an abnormality in the patient based on a Mahalanobis distance between a distribution of a feature group consisting of the first feature amounts for each of the plurality of stable periods and the second feature amount;
outputting the result of said detection;
A program that causes a computer to execute the following:

Images