JP2019097829A - Abnormality notification device and program - Google Patents

Abstract

To provide an abnormality notification device and the like, enabling the abnormality notification accuracy thereof to be improved by presuming a future condition of an object person being a patient depending on a personal characteristic and daily rhythm of the object person.SOLUTION: The abnormality notification device comprises: biological signal acquisition means acquiring a biological signal of an object person on a bed; biological information value calculation means calculating a biological information value using the acquired biological signal; presumption means presuming a condition of the object person after the lapse of a presumption time on the basis of the biological information value; notification means performing notification when it is determined by the presumption means that the condition of the object person is abnormal; and condition determination means determining the current condition of the present object person on the basis of the biological information value. On the basis of a result learned from the current condition of the present object person determined by the condition determination means, the presumption means presumes the condition of the object person.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an abnormality notification device and the like.

  2. Description of the Related Art There are conventionally known devices and systems that report patient abnormalities. For example, as disclosed in Patent Document 1, the living activity and life activity of the subject are detected by the noninvasive vital sensor and classified into a plurality, and the allowable continuation time for each classification is sequentially integrated, and the integration time exceeds the threshold The invention of reporting is known.

Patent No. 3557777 gazette

  Conventionally, it is common to make a notification depending on whether a measured value (biometric information value) associated with an abnormality of a subject exceeds a threshold. For example, in Patent Document 1 described above as well, regardless of the strength of the relationship between the subject person's condition and the abnormality, it is determined that the life activity or the life activity accumulated time exceeds the predetermined threshold, and notification is given. However, the accuracy decreases depending on the condition of the subject, and the strength of the association between life activities and life activities and abnormalities fluctuates depending on conditions. Therefore, since the notification is simply made regardless of the level of the relationship between the accuracy fluctuation and the abnormality, it is reported without reliability under the condition that the relationship between the condition of the target person whose accuracy decreases and the abnormality becomes weak. Had a problem of

  In addition, when the normal range of strong biological information values related to the condition of the subject such as heart rate and respiratory rate is set, and it is determined as abnormal if it exceeds the normal range, it does not deviate from the normal range. Abnormalities may be missed, or the heart rate may increase temporarily due to exercise, so it may be erroneously judged as abnormal, or an athlete with high cardiopulmonary function may be erroneously judged as abnormal because of a low heart rate even in normal times or even in abnormal times It may not deviate from the normal range. As described above, when the abnormality is determined only from the deviation of the normal range, there is a problem that in the alert indicating the abnormality, the false alarm or the omission of the abnormality is likely to be many. Not only individual differences exist in the normal range, but the heart rate and respiratory rate also change with time because they show high daily rhythm and low daily rhythm at night. In order to remove the influence of transient outliers caused by contamination of an artifact and an artifact unrelated to an abnormality, there is a method of analyzing a long-term change and reporting an abnormality. However, in the method of capturing changes from long-term data, it is necessary to unify and analyze data conditions. For example, in data in which exercise and resting are mixed, if the change is not analyzed after the two are separated, the accuracy of the abnormal notification is lowered.

  In particular, in the case of a system that reports an abnormality based on biological information values used in a hospital or a nursing home, unnecessary abnormality notification based on an error or the like causes unnecessary burden on the medical staff and staff. The problem is that missing the anomaly causes a fatal situation.

  Furthermore, there is a problem that planned responses can not be made after departure from the normal range, and there is a problem that the prognosis will deteriorate. Therefore, the future state of the patient is estimated to some extent, and any abnormality that may occur in the future based on the estimation result. There is a need to predict and to notify in advance. However, such prediction of the future is even more difficult, resulting in a notification with many false alarms.

  In view of the problems described above, according to the present invention, it is possible to improve the accuracy of inferring an abnormality by inferring the future state according to the personal characteristics and circadian rhythm of a subject who is a patient It is providing an abnormality alerting device etc.

  The abnormality determination apparatus according to the present invention is based on biological signal acquisition means for acquiring a biological signal in a bed of a subject, biological information value calculation means for calculating a biological information value from the acquired biological signal, and the biological information value. And estimating means for estimating the state of the subject after an estimation time has elapsed, notifying means for notifying when the state of the subject is determined to be abnormal by the estimating means, and based on the biological information value. And state determination means for determining the current state of the target person, wherein the estimation means is based on the result of learning from the current state of the target person determined by the state determination means. It is characterized by inferring a state.

  The program according to the present invention includes a computer, a biological signal acquisition function of acquiring a biological signal in a bed of a subject, a biological information value calculation function of calculating a biological information value from the acquired biological signal, and the biological information value. Based on the estimation function for estimating the condition of the subject after the estimation time has elapsed, the notification function for notifying when the condition of the subject is determined to be abnormal by the estimation function, and the biological information value. Program for realizing the state determination function of determining the current state of the target person, wherein the inference function is based on the result of learning from the current state of the target person determined by the state determination function. And estimating the state of the subject.

  According to the present invention, it is possible to improve the accuracy of inferring and reporting an abnormal state according to the personal characteristics of the subject who is a patient and the circadian rhythm.

It is a figure for demonstrating the whole in 1st Embodiment. It is a figure for demonstrating the function structure in 1st Embodiment. It is an operation | movement flow for demonstrating the patient estimation process in 1st Embodiment. It is a figure for demonstrating the neural network in 2nd Embodiment. It is a figure for demonstrating the recurrent neural network in 2nd Embodiment. It is a figure for demonstrating the function structure in 3rd Embodiment. It is a figure for demonstrating the operation | movement of the patient estimation part in 4th Embodiment. It is a figure for demonstrating the whole system in 5th Embodiment.

  Hereinafter, one embodiment for carrying out the present invention will be described with reference to the drawings. Although the case where the abnormality alerting | reporting apparatus of this invention is specifically applied is demonstrated, the range to which this invention is applied is not limited to the said embodiment.

[1. First embodiment]
[1.1 whole system]
FIG. 1 is a diagram for explaining the overall outline of an abnormality notification system 1 to which the present invention is applied. As shown in FIG. 1, the abnormality notification system 1 includes a processing unit 5 for processing a value output from the detection device 3 placed between the floor of the bed 10 and the mattress 20 and the detection device 3 It is configured to be equipped. The detection device 3 and the processing device 5 constitute an output device of biological information.

  When a target person (hereinafter, referred to as “patient P” as an example) is placed on the mattress 20, the detection device 3 detects body vibration (vibration generated from the human body) as a biological signal of the patient P who is the target person. Then, based on the detected vibration, the biological information value of the patient P is calculated. In the present embodiment, the calculated biological information value (at least the respiration rate, the heart rate, and the amount of activity) can be output and displayed as the biological information value of the patient P. For example, the detection device 3 may be integrally formed by providing a storage unit, a display unit, and the like. In addition, since the processing device 5 may be a general-purpose device, the processing device 5 is not limited to an information processing device such as a computer, and may be configured by a device such as a tablet or a smartphone.

  In addition, the target person may be a person being treated for illness, or one requiring care. In addition, even if it is a healthy person who does not need care, it may be an elderly person, a child, a disabled person, or a person or animal.

  Here, the detection device 3 is configured in a sheet shape so as to be thinner. As a result, even if the patient P is placed between the bed 10 and the mattress 20, the patient P can be used without feeling discomfort, so the biological information value on the bed is measured over a long period of time (for example, 1 hour or more, 8) It is possible to have a predetermined period such as time or more, or a predetermined period such as one night, one sleep, one week, one month, one year, ten years or more. That is, since the biological information value is calculated from the body vibration, the respiratory rate and the heart rate can not be measured when the subject is moving the body (due to the measurement accuracy of the respiratory rate and the heart rate being lowered during body movement, abnormality It becomes a noise for the notification system), and acquires a biological information value etc. as a state of the patient limited at rest. Furthermore, the detection device 3 has a function of determining the reliability of the measured body vibration data and recording only highly reliable data.

  In addition, the detection apparatus 3 should just be able to acquire the patient's P biological signal (a body movement, a respiratory movement, a heart beat etc.). In the present embodiment, although the heart rate and the respiration rate are calculated based on the body vibration, for example, detection is performed using an infrared sensor, or a biological signal of the patient P is acquired by an acquired image or the like, distortion A gauged actuator may be used. In addition, by using a built-in acceleration sensor or the like, it may be realized by, for example, a smartphone or a tablet placed on the bed 10.

[1.2 Functional configuration]
Subsequently, the functional configuration of the abnormality notification system 1 will be described with reference to FIG. The abnormality notification system 1 in the present embodiment is configured to include the detection device 3 and the processing device 5, and each functional unit (process) may be realized by any means other than the biological signal acquisition unit 200. . That is, by combining these devices, the device functions as an abnormality notification device.

  Note that the abnormality notification system 1 may be a staff or a family to which an abnormality is reported. Further, as a method of making a notification, notification (notification) may be made simply by sound or screen display, or notification may be made to the portable terminal device by e-mail or the like. In addition, notification (notification) may be made to other terminal devices and the like.

  The abnormality notification system 1 (abnormality notification device) includes a control unit 100, a biological signal acquisition unit 200, a biological information value calculation unit 300, a sleep state determination unit 350, an input unit 400, an output unit 450, and a storage unit. It comprises 500, a patient state acquisition unit 600, a patient state estimation unit 700, a learning unit 800, and an alert output unit 900. In the case of FIG. 1, the control unit 100, the biological signal acquisition unit 200, and the storage unit 500 are included in the detection device 3, and the processing device 5 is otherwise included. The patient state acquisition unit 600 may use the biological signal acquisition unit 200 or may be separately provided in the bed 10.

  The control unit 100 is a functional unit for controlling the operation of the abnormality notification system 1. For example, it may be configured by a control device such as a CPU or may be configured by a control device such as a computer. The control unit 100 realizes various processes by reading and executing various programs stored in the storage unit 500. In the present embodiment, the control unit 100 operates as a whole, but may be provided in each of the detection device 3 and the processing device 5.

  The biological signal acquisition unit 200 is a functional unit for acquiring a biological signal of the patient P. In the present embodiment, as an example, a sensor that detects a pressure change is used to acquire body vibration which is a type of biological signal. And the acquired body vibration is converted into biometric information value data, such as a respiration rate, a heart rate, and an amount of activity, and is output. Furthermore, based on the body vibration data, the patient's bed rest state (for example, whether the patient P is bed rest or not, bed rest, bed rest, end sitting position, etc.) or sleep state as described later (sleep, awakening) It is also possible to obtain

  Although the biological signal acquisition unit 200 in the present embodiment acquires body vibration of the patient by using a pressure sensor and acquires respiration and heart rate from body vibration, for example, the load sensor changes the position of the center of gravity of the patient and the load value. Alternatively, the biological signal may be acquired, or by providing a microphone, the biological signal may be acquired based on the sound picked up by the microphone. It is sufficient if any of the sensors can be used to obtain the patient's biological signal.

  That is, the biological signal acquisition unit 200 may be connected to a device such as the detection device 3 or may be configured to receive a biological signal from an external device.

  The biological information value calculation unit 300 is a functional unit for calculating biological information values (respiration rate, heart rate, and the like) of the patient P. In the present embodiment, the respiration component and the heart rate component may be extracted from the body movement acquired from the biological signal acquisition unit 200, and the respiration rate and the heart rate may be determined based on the respiration interval and the heart interval. In addition, periodicity of body movement may be analyzed (Fourier transform etc.) to calculate respiratory rate and heart rate from peak frequency, or pattern recognition or artificial intelligence (machine learning) may be used. .

  The sleep state determination unit 350 is a functional unit for determining the sleep state of the patient. For example, based on the biological signal acquired by the biological information value calculation unit 300, the sleep state of the patient is determined. The sleep state may be determined as "wake up" or "sleep", and further, sleep may be determined as "REM sleep" or "non-REM sleep", or the depth of sleep may be determined.

  The input unit 400 is a functional unit for a measurer to input various conditions or to input an operation to start measurement. For example, it is realized by any input means such as a hardware key or a software key.

  The output unit 450 is a functional unit for outputting biological information values such as the sleep state, the heart rate, and the respiratory rate, and for reporting an abnormality. The output unit 450 may be a display device such as a display or a notification device (sound output device) for notifying an alarm or the like. Also, it may be an external storage device that stores data, a transmission device that transmits data via a communication channel, or the like. In addition, it may be a communication device in the case of reporting to another device.

  The storage unit 500 is a functional unit that stores various data and programs for operating the abnormality notification system 1. The control unit 100 realizes various functions by reading and executing the program stored in the storage unit 500. Here, the storage unit 500 is configured of, for example, a semiconductor memory, a magnetic disk device, or the like. Here, in the storage unit 500, biometric information data 510 is stored.

  The biological information data 510 stores the respiration rate and the heart rate obtained from the acquired biological signal (body movement). In the present embodiment, although the respiration rate, the heart rate, and the body movement are stored, at least one of them may be stored. In addition, if it is a biological information value that can be calculated by the biological information value calculation unit 300, other information (for example, the disordered breathing index based on the fluctuation of the breathing amplitude, the periodic physical activity index based on the periodicity of body movement) You may memorize.

  The sleep state data 520 stores the sleep state of the patient. As the sleep state determined by the sleep state determination unit 350, states such as “sleep” and “wake up” and “in bed” and “getting out of bed” acquired by the patient state acquisition unit 600 are stored.

  The patient state acquisition unit 600 is a functional unit for acquiring the state of the patient. For example, the load sensor or the like provided on the bed 10 acquires the condition (uplifting, staying, etc.) of the patient. As described above, it may be realized in the biological signal acquisition unit 200.

  The patient state estimation unit 700 is a functional unit for estimating a patient state from parameters such as a biological information value. When the patient state estimation unit 700 estimates that the condition of the patient is abnormal, the alert output unit 900 outputs (informs) an alert.

  Here, a method in which the patient state in the patient state estimation unit 700 in the present embodiment is estimated to be abnormal will be described.

  Here, a case where the patient state estimation unit 700 estimates a patient state using artificial intelligence (machine learning) will be described. The patient state estimation unit 700 uses biological information and the patient state as input values (input data), and estimates the patient state by using artificial intelligence and various statistical indicators.

  As shown in FIG. 3, the patient state estimation unit 700 includes a feature extraction unit 710, an identification unit 720, an identification dictionary 730, and a patient state output unit 740.

  First, various parameters are input and used as input data input to the patient state estimation unit 700. For example, in the present embodiment, “breathing rate”, “heart rate”, “sleeping state”, and “activity amount” calculated from the body vibration data acquired by the biological signal acquiring unit 200 are used. It is also possible to use “respiratory rate variation” and “heart rate variation” calculated from these biological information values, and “respiratory disorder index” and “periodic physical activity index” calculated from the same body vibration data.

  Here, the "sleeping state" includes the states of "living in the bed" and the state of "going up", and when in the bed, the states of "wake up" and "sleeping" can be specified. "Sleep" may be further classified as "REM sleep" or "non-REM sleep", or the depth of sleep may be determined. In addition, as “respiratory disorder index”, although the number of significant fluctuation of respiratory amplitude per hour of sleep is used, the number of apnea per hour of sleep (apnea index) or The total number of apnea and hypopnea (apnea-hypopnea index) may be used. In addition, “periodic body movement index” uses the number of occurrences of periodic body movement per hour of sleep, but may use the number of periodic limb movements per hour of sleep.

  Then, the feature extraction unit 710 extracts each feature point and outputs it as a feature vector. Here, for example, the following can be considered to be extracted as feature points.

(1) Respiration rate of 30 [times / minute] or more or 8 [times / minute] or less continues for a fixed time or more (2) Heart rate 120 [times / minute] or more or 40 [times / minute] or less continues for a fixed time or more (3) The trend of heart rate or breathing rate rises (10% or more) from the start to the end of night sleep
(4) Variability (standard deviation, coefficient of variation) of respiratory rate or heart rate at night (21:00 to 6:59) is more than a fixed value (5) Breathing disorder index or periodicity movement index decreases significantly (6 ) Dyspnea index or periodic movement index increase significantly or more than a certain value (night)
(7) Activity amount increases or decreases significantly (8) Sleep determination continues for a fixed time or more, and awakening determination at night is 95% or more

  A feature vector is output by combining one or more of these feature points. In addition, what was demonstrated as a feature point is one example, and is not limited to the said value. For example, taking (1) as an example, the respiration rate may be 25 [times / minute] or more, or 10 [times / minute] or less. Thus, each value is for convenience of explanation. Then, the corresponding feature point may be “1”, and the non-corresponding feature point may be “0”, or a random variable may be output.

  When all the feature points described above are included, the feature space is eight-dimensional and is output to the identification unit 720 as an eight-dimensional feature vector.

  The identification unit 720 identifies a class corresponding to a patient state from the input feature vector. At this time, the class is identified by matching with a plurality of prototypes prepared in advance as an identification dictionary 730. The prototype may be stored as a feature vector corresponding to each class, or may store a feature vector representing the class.

  If a feature vector representing a class is stored, the class to which the closest prototype belongs is determined. At this time, it may be determined by the nearest neighbor determination rule or may be identified by the k-nearest neighbor method.

  Although the identification dictionary 730 used by the identification unit 720 may store a prototype in advance, in the present embodiment, the identification dictionary 730 is updated by using machine learning by the learning unit 800 described later.

  Then, the patient state output unit 740 outputs the patient state corresponding to the class identified by the identification unit 720. The state of the patient to be output may be “normal” or “abnormal”, and “abnormal” may be identified as “fever”, “capacitance change” or the like, or a random variable may be output.

  Thus, according to the present embodiment, biological information including “respiratory rate”, “heart rate”, “activity amount”, “getting out of bed”, and “being in bed” is acquired, and the state of the patient is estimated from these pieces of biological information. It becomes possible.

  That is, as the identification dictionary 730, the feature vector described above and the patient state are stored, and a patient state that matches the feature vector is determined based on the identification dictionary 730. In the embodiment described above, the feature vector and the patient state are stored in the identification dictionary 730. However, for example, a biological information value or a patient state such as a sleep state may be stored directly. In addition, the state of those changes (for example, an increase in heart rate, a decrease in respiratory rate, etc.) and the estimated state of the patient may be stored in association with each other.

  The condition of the patient may be the condition of the patient after a predetermined time (the condition of the patient to be estimated). For example, the state of the patient after a predetermined time has been stored, such as after 30 minutes, 1 hour, 2 hours,. In addition, after the predetermined time has elapsed, for example, in the case of the morning, the patient's state after a predetermined time has elapsed, such as a time interval of night, light off time, sleep start time, bedtime, wakeup time, and bedtime elapsed time. It may be stored. Also, a plurality of these may be stored, which may be stored in association with the current patient state. For example, based on the current biological information value and changes in the biological information value, the current patient's condition, the patient's condition after one hour, the patient's condition the next morning, and the transition of the patient's condition for the next 24 hours By using machine learning, it is possible to guess.

  The learning unit 800 can improve the accuracy of estimating (predicting) the state of the patient by performing machine learning on the identification dictionary 730. Here, in order to generate correct data (teacher data), the following processing is performed.

  First, the condition of the patient is abnormal from the respiratory rate at the determination time which is the time to execute the process of determining the correct data (teacher data), the heart rate, the sleep awakening determination, the activity amount, the respiratory disorder index, and the periodic activity index Predict what it will be.

Specifically, within the judgment elapsed time (for example, within 1 hour to 12 hours, preferably within 8 hours) from the judgment time,
(1) Respiration rate of 30 [times / minute] or more continues continuously for 60 minutes or more (tachypnea continues for a certain time or more)
(2) The respiratory rate drops by 10 [times / minute] or more and the heart rate drops by 20 [times / minute] or less within 60 minutes in the trend (the respiratory rate and the heart rate decrease with each change in a fixed time) about)
(3) In patients who are not supposed to go out of bed, stay-out judgment continues for 3 minutes or more (in patients who can not go out of bed, stay-out judgment continues for a certain period of time or more)
It predicts that the condition (abnormal condition) which applies to the condition will occur.

  Here, learning is performed by comparing the above prediction result with whether or not an abnormal state has actually occurred. For example, after an abnormal state is predicted to occur in the above prediction result, it is determined that the prediction was correct when an abnormal state occurs within the determination elapsed time. That is, information at the time of this prediction (for example, the above-mentioned respiration rate, heart rate, sleep awakening determination, activity amount, disordered breathing index, corrected body movement index) is learned as correct data. On the other hand, even if it is predicted that an abnormal condition will occur, if no abnormality occurs within the judgment elapsed time, it is learned that the predicted result is not correct.

  Similarly, in the above prediction result, after it is predicted that no abnormal condition will occur, it is determined that the prediction was correct if no abnormal condition has occurred within the judgment elapsed time. That is, information at the time of this prediction (for example, the above-mentioned respiration rate, heart rate, sleep awakening determination, activity amount, disordered breathing index, corrected body movement index) is learned as correct data. On the other hand, even if it is predicted that no abnormal condition will occur, if an abnormality occurs within the judgment elapsed time, it is learned that the prediction result is not correct.

  As described above, by repeatedly determining the patient state estimated by the patient state estimation unit 700 and the patient state determined by the learning unit 800, it is possible to estimate the patient with higher accuracy. become.

  In the present embodiment, the conditions for generating the correct data for learning are described as an example. However, learning may be performed using a state in which the patient's state is actually inferred. Also in this case, for example, after the estimation time has elapsed, additional learning is performed with the actual patient state as correct data (teacher data) by comparing the estimated patient state with the actual patient state. Is possible.

[2. Second embodiment]
Subsequently, the second embodiment will be described. The second embodiment differs from the first embodiment in that the patient state estimation unit 700 uses a neural network.

  First, the patient state estimation unit 700 receives and uses the various parameters described above. For example, in the present embodiment, “breathing rate”, “heart rate”, “sleeping state”, and “activity amount” calculated from the body vibration data acquired by the biological signal acquiring unit 200 are used. It is also possible to use “respiratory rate variation” and “heart rate variation” calculated from these biological information values, and “respiratory disorder index” and “periodic physical activity index” calculated from the same body vibration data. In addition, sleep states such as “bed up” including “sleep” and “wake up” of the patient, and “bed up” are also available.

  These biological information values, sleep states, and other patient states (hereinafter referred to as "patient biological information") are input to a neural network composed of a plurality of layers and neurons included in each layer. Each neuron receives a signal from another plurality of neurons, and outputs the calculated signal to another plurality of neurons. When the neural network has a multilayer structure, it is referred to as an input layer, an intermediate layer (hidden layer), and an output layer in the order in which the signals flow.

  If the middle layer of the neural network consists of multiple layers, it is called a deep neural network (for example, Convolutional Neural Network with convolution operation), and the machine learning method using this is called deep learning. Call.

  Patient biological information is subjected to various operations (convolution operation, pooling operation, normalization operation, matrix operation, etc.) on neurons in each layer of the neural network, flows while changing its form, and a plurality of signals are output from the output layer.

  Each of the plurality of output values from the neural network is associated with the condition of the patient, and processing is performed such as inferring the condition of the patient associated with the output value having the largest value. Alternatively, one or more output values may be passed through a classifier without directly outputting the patient's condition, and the patient's condition may be inferred from the output of the classifier.

  Parameters that are coefficients used for various operations of the neural network are input in advance to the neural network a large number of patient biological information and the patient's state of the patient biological information, and the error between the output value and the correct value is Propagation is determined by propagating the neural network backwards and updating the parameters of the neurons in each layer many times. The process of updating and determining parameters in this manner is called learning.

  The structure of the neural network and individual operations are known techniques described in a book or a paper, and any one of them may be used.

  As described above, by using the patient state estimation unit 700, the patient state is output from input data such as patient's biological information.

  Then, the learning unit 800 corrects data (teacher data) based on the patient's estimation result output from the patient condition estimation unit 700 and the actual patient condition acquired and determined by the biological signal acquisition unit 200. Do learning based on

  In this case, learning is performed by adjusting the weighting factor between the input layer, the intermediate layer, and the output layer in the neural network. In addition, it is possible to reduce the error from the correct data (teacher data) by learning, but at this time, at the beginning, the learning rate is increased and weights are frequently updated, and the learning time According to the method, the learning rate may be reduced and the weight may be finely adjusted.

  Further, it is possible to use an inverse error propagation method in which the weight is adjusted so as to approximate the correct result by taking an error between the patient's estimation result and the patient's condition and propagating this error to each layer.

  In addition, the learning unit 800 can improve the accuracy of the neural network by learning. Here, in order to generate correct answer data (teacher data), the same process as that of the first embodiment is executed.

  First, the condition of the patient is abnormal from the respiratory rate at the determination time which is the time to execute the process of determining the correct data (teacher data), the heart rate, the sleep awakening determination, the activity amount, the respiratory disorder index, and the periodic activity index Predict what it will be.

Specifically, within the judgment elapsed time (for example, within 1 hour to 12 hours, preferably within 8 hours) from the judgment time,
(1) Respiration rate of 30 [times / minute] or more continues continuously for 60 minutes or more (tachypnea continues for a certain time or more)
(2) The respiratory rate drops by 10 [times / minute] or more and the heart rate drops by 20 [times / minute] or less within 60 minutes in the trend (the respiratory rate and the heart rate decrease with each change in a fixed time) about)
(3) In patients who are not supposed to go out of bed, stay-out judgment continues for 3 minutes or more (in patients who can not go out of bed, stay-out judgment continues for a certain period of time or more)
It predicts that the condition (abnormal condition) which applies to the condition will occur.

  Here, learning is performed by comparing the above prediction result with whether or not an abnormal state has actually occurred. For example, after an abnormal state is predicted to occur in the above prediction result, it is determined that the prediction was correct when an abnormal state occurs within the determination elapsed time. That is, information at the time of this prediction (for example, the above-mentioned respiration rate, heart rate, sleep awakening determination, activity amount, disordered breathing index, corrected body movement index) is learned as correct data. On the other hand, even if it is predicted that an abnormal condition will occur, if no abnormality occurs within the judgment elapsed time, it is learned that the predicted result is not correct.

  Similarly, in the above prediction result, after it is predicted that no abnormal condition will occur, it is determined that the prediction was correct if no abnormal condition has occurred within the judgment elapsed time. That is, information at the time of this prediction (for example, the above-mentioned respiration rate, heart rate, sleep awakening determination, activity amount, disordered breathing index, corrected body movement index) is learned as correct data. On the other hand, even if it is predicted that no abnormal condition will occur, if an abnormality occurs within the judgment elapsed time, it is learned that the prediction result is not correct.

  Further, in the case of biological information values calculated based on biological signals, or in the case of time-series data such as sleep, a Recurrent Neural Network (RNN) may be used. This is to extend the method of the neural network as described above so that time series data can be handled. As recurrent neural networks, there are various networks such as Elman Network (Elman Network), Jordan Network (Jordan Network), Echo State Network (Echo State Network), and Long Short-Term Memory Network (LSTM). The network will be described as an example.

  For example, as shown in FIG. 5, in the Elman network, not only data at time t but also data of a hidden layer (intermediate layer) at time t-1 can be used. With such a network configuration, past patient biometric information can be influenced by the current prediction, and it is possible to estimate the patient's state based on the relationship in time.

  As described above, according to the present embodiment, patient biological information (for example, "respiratory rate" "heart rate" "sleeping state" "activity amount" "variation in respiratory rate" "variation in heart rate" "respiratory disorder index" Appropriately infer the patient's condition by using neural networks, recurrent networks, etc. from various information such as “periodic physical activity index”, “sleeping”, “wakening”, “bedding”, “bedding”, etc. Will be able to

  In addition, by utilizing the state of the patient who was actually freed as correct data (teacher data), without causing the user's hand (for example, medical staff, facility staff, assistants, etc.) It is possible to improve the estimation accuracy of the patient condition.

[3. Third embodiment]
Subsequently, the third embodiment will be described. In the third embodiment, the functional configuration of FIG. 2 of the first embodiment is replaced with FIG.

  In addition to the functional configuration of the first embodiment, a patient diary output unit 650 is further provided. Further, instead of the patient state estimation unit 700, a patient state estimation unit 750 is provided which estimates the patient state using a neural network based on image data of a patient diary.

  The patient diary output unit 650 calculates the acquired biological information value, and the sleep state (0: getting out of bed, 1: being in bed, awakening, 2: being sleep) values of pixel values per minute, where one row represents 24 hours. It is a functional unit that outputs as image data (image data of “1440 pixels × pixels for days”). The patient diary includes a respiration diary representing the patient's respiration rate, a heart rate diary representing the patient's heart rate, a sleep diary representing the patient's sleep state, an activity amount diary representing the patient's movement, and a respiratory disorder representing the number of respiratory disorder events A diary, periodic movement diary representing the number of periodic movement events, etc. can be output. Note that these parameters may be combined and output as one patient diary. The graphs of these patient diaries can be output as diary data which is image data.

  The patient state estimation unit 700 is a functional unit for estimating a patient state from the input diary data. Here, as a process of estimating a patient state, recently, deep learning (deep neural network) has given high accuracy particularly in image recognition, and this method is used as an example in this embodiment. In addition, about the process in deep learning, since it demonstrated by embodiment mentioned above, detailed description is abbreviate | omitted.

  Parameters which are coefficients used for various operations of the neural network in the present embodiment input in advance many diary data and the patient's condition of the diary data to the neural network, and the error between the output value and the correct value is The error back propagation method is determined by propagating the neural network backwards and updating the parameters of the neurons of each layer many times. The process of updating and determining parameters in this manner is called learning.

  The structure of the neural network and individual operations are known techniques described in a book or a paper, and any one of them may be used.

  By using the patient state estimation unit 750, the patient state is output from input data such as patient's biological information.

  In the embodiment described above, a neural network is used by inputting diary data in which one line is 24 hours, but in consideration of the rhythmicity on a weekly basis, one line is 7 days, and the diary data is roughly a month. It is good as diary data which made one line 28 days in consideration of rhythmicity of unit, diary data which made one line 365 days in consideration of annual rhythmicity, etc., and a living body which does not consider rhythmicity beforehand Information values may be input to use a neural network. That is, the patient state may be estimated by inputting information such as "heart rate," "breathing rate," "activity amount," "getting out of bed," and "at bedtime" in a neural network with each time axis synchronized. .

  Also in the present embodiment, it is possible to learn about the patient state estimation unit 750 by using the learning unit 800.

  That is, learning is performed based on the patient state estimated by the patient state estimation unit 750 and the patient state determined from the patient biological information calculated and determined based on the biological signal acquisition unit 200. Do.

  That is, learning is performed using diary data as correct data (teacher data) when it is estimated from the diary data that the patient's condition is abnormal and the patient's condition is actually determined to be abnormal. In addition, when it is presumed that the patient's condition is not abnormal and thereafter the patient's condition is not judged to be abnormal, learning may be performed with the diary data at that time as correct data.

  As described above, according to the present embodiment, it is possible to estimate the patient's condition by using the diary data without using detailed patient biometric information.

[4. Fourth embodiment]
Subsequently, the fourth embodiment will be described. In the embodiment described above, the embodiment in which the state of the patient is inferred using machine learning or a neural network has been described. In this embodiment, multivariate analysis is used to infer the condition of a patient.

  For example, multivariate analysis using the measured data up to now as the dependent variable, with the respiratory rate / heart rate above or below a certain level or sudden drop, which is a feature that is uniformly recognized during death, critically ill and acute phase Use to estimate the patient's condition.

  Here, a method in which the patient state in the patient state estimation unit 700 in the present embodiment is estimated to be abnormal will be described.

  FIG. 7 is an operation flow for describing a patient state estimation process for estimating a patient state. In the present embodiment, the patient state estimation unit 700 estimates the patient state by executing the patient state estimation process of FIG. 7.

  First, a biological information value is obtained (calculated) (step S102). Here, as vital information values, respiratory rate, heart rate, and activity amount are important, but it is not possible to get sleep by acquiring the patient's condition such as sleep, awakening (staying), getting out of bed, time in bed It becomes possible to estimate patient's condition in more detail, taking into consideration changes such as increased time not spent in bed, continuous residence time and continuous bedtime, etc.

  Furthermore, as one of the biological information values, the patient's index (biometric index), respiratory disorder index, periodic physical activity index by acquiring the value of these absolute values, change in daily average value, 24 hours More detailed patient state estimation can be made from changes in time-series distribution, etc. In addition, the history of the biological information value may be acquired, and the past value, the average value, the standard deviation, the variation coefficient, and the value / proportion of the change of the latest predetermined time may be acquired.

  The biological information value may be directly obtained as a biological information value, or may be calculated by performing a predetermined operation from a biological signal. Further, another biometric information value or an index may be calculated from one biometric information value.

  Subsequently, it is determined whether or not the abnormality determination condition is met (step S104). If the abnormality determination condition is met, one is added to the number of abnormality determinations (step S104; Yes → step S106). Then, if the determination is not completed for all the abnormality determination conditions, the next abnormality determination condition is read out and it is similarly determined whether the abnormality determination conditions are met (Step S108; No → Step S110 → Step S104).

  That is, when the patient state estimation unit 700 estimates a patient state, it is determined based on the biological information value and the sleep state whether or not the plurality of abnormality determination conditions are met. Here, an example of the abnormality determination condition will be described below.

That is, as the abnormality determination condition, the following conditions can be considered.
(1) Average respiratory rate for the last 30 minutes (The accuracy is improved by using a value for a relatively long time rather than an instantaneous value)
(2) The difference between the latest nightly average respiratory rate and the past average value (The average nightly respiratory rate has little variation within the individual and the accuracy is good)
(3) Average heart rate for the last 60 minutes (Because measurement accuracy is lower than respiration rate, calculation time is longer than (1))
(4) The difference between the latest average value of the average heart rate at night and the past average value (the measurement accuracy is lower than the respiration rate, so the abnormality determination condition is less likely to be met than (2), or the weight of the abnormality determination result is small Do)
(5) Slope of linear approximation line of respiratory rate at night (High accuracy due to global fluctuation tendency. Respiratory rate tends to rise from night to morning, such as the difference between the average value in the first half of the night and the average value in the second half It is sufficient if it is an indicator that can assess whether there is a downtrend or not.)
(6) Slope of linear approximation line of heart rate at night (it is high accuracy due to global fluctuation tendency but lower accuracy than respiration rate, so it is difficult to satisfy abnormality judgment condition or weight of abnormality judgment result It should be smaller, as long as it is an indicator that can assess whether the heart rate is rising or falling from night to morning, such as the difference between the average value in the first half of the night and the average in the second half.)
(7) Variability in the nighttime respiratory rate (Individual-specific indicator, high accuracy due to global fluctuation tendency. Standard deviation, coefficient of variation, etc.)
(8) Heart rate variability at night (this is an index unique to individuals, high accuracy due to global fluctuation tendency but lower accuracy than respiration rate, so it is difficult to meet the abnormality judgment condition, or abnormality judgment result Reduce the weight of (standard deviation, coefficient of variation, etc.)
(9) The difference between the latest nightly average activity amount and the past average value (this is an individual-specific indicator, and its accuracy is high due to global fluctuations)
(10) The difference between the latest nightly average respiratory disorder index and the past average (It is an individual-specific indicator, and its accuracy is high due to the global fluctuation tendency)
(11) The difference between the latest and the average value of the nightly average periodic physical activity index (It is an individual-specific index, and its accuracy is high due to the global fluctuation tendency)
(12) The difference between the latest nightly average take-away time and the past average (It is an individual-specific indicator, and its accuracy is high due to global fluctuations.)
(13) Integrated value of the difference between the average occupancy rate (0 to 1) for 24 hours (every 1 minute) and the latest 24 hours judgment (presence: 1, bed departure: 0) -144) (Individual-specific indicator, high accuracy due to global fluctuation tendency)
(14) Average activity amount over the last 8 hours (The relationship with activity is a strong indicator, and its accuracy is high due to global fluctuation tendency)

  It is determined based on each of these abnormality determination conditions whether or not the reference value is exceeded. For example, in the case of the abnormality determination condition (1), the determination is performed using the respiration rate among the input biological information. For example, the average respiration rate for the last 30 minutes is calculated, and if the average respiration rate does not fall within the reference value (for example, 8 to 28), it is determined as abnormal and one is added to the abnormality determination number.

  In each anomaly judgment condition, the method of deciding whether the patient's condition is abnormal or not is missed by the target person's attribute, current disease or medical history, characteristics of the biosignal acquisition unit, false alarm or not May be changed as appropriate depending on whether you want to For example, when responding to a medical history, it may be considered that the patient has a heart condition and if it should be noted, the importance is increased. In addition, when the heart has a chronic disease and an arrhythmia is produced, the accuracy of the heart rate may be lowered to lower the importance of the condition related to the heart rate.

  Also, for example, as the characteristics of the biological signal acquisition unit 200, it is conceivable to change the weighting depending on the product if there is a difference in accuracy. For example, in the case of being placed under the subject and based on body vibration, the weight of the respiration rate is weighted in order to obtain the respiration rate more accurately. Also, in the case of an electrocardiograph, in order to obtain the heart rate more accurately, the weight of the heart rate is weighted. Thus, the importance (weighting or priority) may be assigned according to the type of sensor.

  That is, it is important to determine a patient's abnormality by combining a plurality of these abnormality determination conditions. For example, when “3” is set as the abnormality reference value, if the number of abnormality judgments is “3” or more which is the abnormality reference value, the state of the patient is judged as “abnormal” (step S112; Yes) → Step S114). If the number is less than that, for example, if the number of abnormality determinations is “2” or less (if the abnormality reference value is “3”), it is determined that the patient's condition is normal.

  In FIG. 7, the number of abnormality determinations and the abnormality reference value are used to simply determine the number of abnormality determination conditions. However, the patient state can be determined by other methods.

  For example, instead of determining the conformity to each abnormality determination condition, the degree of abnormality may be calculated for each of them using the abnormality degree determination formula, and the patient state may be determined using the total value of the calculated degrees of abnormality. For example, multivariate analysis of each value may be performed to calculate the overall degree of abnormality and to determine the patient state.

  Moreover, it is not necessary to use all the abnormality determination conditions, and it is also possible to combine them as needed. In addition, the strength of the association between each abnormality determination condition and the true abnormality is not uniform, and the above-mentioned attributes of the subject, the present disease or medical history, the characteristics of the biological signal acquisition unit, the missing report Because it changes depending on whether you want to reduce) or the like, you may use the number of abnormality judgments weighted according to the strength of association with the true abnormality.

  In addition, among the plurality of abnormality determination conditions, important conditions may be preferentially used. For example, in the case where the user is placed under the subject and is based on body vibration, the condition (1) is the most effective among the above-mentioned abnormality determination conditions, and the condition is preferentially used or important. The abnormality determination may be performed by performing certain weighting.

  At this time, it is possible to adjust the weighting and priority of parameters for performing multivariate analysis by executing the following processing.

  First, the condition of the patient is abnormal from the respiratory rate at the determination time which is the time to execute the process of determining the correct data (teacher data), the heart rate, the sleep awakening determination, the activity amount, the respiratory disorder index, and the periodic activity index Predict what it will be.

Specifically, within the judgment elapsed time (for example, within 1 hour to 12 hours, preferably within 8 hours) from the judgment time,
(1) Breathing rate more than 30 [times / minute] continuously for 60 minutes or more (2) Within 60 minutes in trend, breathing rate is more than 10 [times / minute] and heart rate is less than 20 [times / minute] Decrease (3) In patients who can not go out of bed, it is predicted that a state (abnormal state) will occur that applies to the condition that the bed up judgment continues for 3 minutes or more.

  Here, the adjustment (learning) of the parameters is performed by comparing the above prediction result and whether or not an abnormal state has actually been achieved. For example, after an abnormal state is predicted to occur in the above prediction result, it is determined that the prediction was correct when an abnormal state occurs within the determination elapsed time. That is, the parameter is adjusted (learned) based on the information at the time of this prediction (for example, the above-mentioned respiration rate, heart rate, sleep awakening determination, activity amount, disordered breathing index, corrected body movement index). On the other hand, even if it is predicted that an abnormal condition will occur, if no abnormality occurs within the judgment elapsed time, it is learned that the predicted result is not correct.

  Similarly, in the above prediction result, after it is predicted that no abnormal condition will occur, it is determined that the prediction was correct if no abnormal condition has occurred within the judgment elapsed time. That is, the parameter is adjusted (learned) based on the information at the time of this prediction (for example, the above-mentioned respiration rate, heart rate, sleep awakening determination, activity amount, disordered breathing index, corrected body movement index). On the other hand, even if it is predicted that no abnormal condition will occur, if an abnormality occurs within the judgment elapsed time, it is learned that the prediction result is not correct.

[5. Fifth embodiment]
The fifth embodiment is an embodiment in the case where the above-described learning is implemented using cloud information and the like.

  For example, as shown in FIG. 8, a plurality of abnormality notification systems 1500 are connected to the network NW in a campus (for example, in the same facility, in a hospital, etc.). Each abnormality notification system 1500 is connected to the management server 1000.

  The management server 1000 may infer the state of the patient (for example, whether it is abnormal) based on the patient biometric information of each abnormality notification system 1500. Also, based on the user data of each anomaly notification system 1500, correct data (teacher data) may be identified and used for learning.

  As described above, by using the states of not only one patient but a plurality of patients, a large number of correct answer data can be obtained, and it is possible to estimate a patient with higher accuracy. The correct teacher data referred to here may be, for example, the entire patient, or may be subdivided according to the attributes of individual patients such as age, gender, height, weight, and medical condition.

  Also, each management server 1000 may be connected to an external service providing server 2000 via the network NW2. That is, by using the cloud to estimate the patient's condition, it is possible to collect more patient biometric information and the like, and it becomes possible to estimate the patient's condition with higher accuracy.

  Further, in the above-described embodiment, although the abnormality notification system 1500 is used, it is possible to use a system capable of acquiring simple patient information. For example, in an information device such as a smart phone or a tablet terminal, the state of a patient may be simply acquired using a built-in acceleration sensor or a gyro sensor. By utilizing many of these simple devices, it is possible to acquire more patient biometric information. Moreover, by using such a simple device, it is not necessary to be necessarily a sick person etc., and a healthy person, a healthy person etc. may use it.

[6. effect]
Thus, according to the above-described embodiment, the respiratory rate and heart rate measured on the bed during the sleeping time (from bedtime to wake-up time) and at night (constant time zone such as 23:00 to 5:59) By using the body movement (activity amount), it is possible to perform highly accurate anomaly notification by acquiring biological information values strongly associated with the anomaly under a unified condition every day.

  In other words, it is possible to capture changes from long-term data accurately by being able to continuously acquire respiratory rate, heart rate, and body movement, which are strongly associated with abnormalities, continuously on a daily basis under unified conditions. Abnormality notification can be performed without being affected. In addition, since there is body movement even at sleep time or at night, by including body movement in the analysis item, it becomes possible to notify abnormality notification taking into consideration fluctuations in heart rate and respiration rate due to body movement artefact and deterioration in accuracy.

  Furthermore, if anomalous notification is given, it will be predicted / estimated whether the patient's condition will become abnormal. By learning based on this predicted / estimated result and the actual patient's condition, more accuracy is achieved. It is possible to estimate the patient's condition as high as possible, and it is possible to report more appropriate abnormalities.

  That is, the patient's biological information obtained from the biological signal acquisition unit 200, the biological information value calculation unit 300, and the sleep state determination unit 350 is not given from the outside (correct data, teacher data) as correct data to be given as learning. , Determine whether the patient was in an abnormal state. Then, by using this determination result, the patient state estimation unit 700 (750) and the learning unit 800 can estimate the patient state with higher accuracy.

[7. Modified example]
The embodiment of the present invention has been described in detail with reference to the drawings, but the specific configuration is not limited to this embodiment, and the design and the like within the scope of the present invention are also claimed. include.

  Moreover, in the present embodiment, the biological information is output in the processing device 5 based on the result output by the detection device 3, but all may be calculated by the detection device 3. In addition to installing the application in a terminal device (for example, a smartphone, a tablet, or a computer) for implementation, for example, processing may be performed on the server side, and the processing result may be returned to the terminal device.

  For example, the processing described above may be realized on the server side by uploading biological information from the detection device 3 to the server. The detection device 3 may be realized by, for example, a device such as a smartphone incorporating an acceleration sensor and a vibration sensor.

  In the embodiment, a program that operates in each device is a program (a program that causes a computer to function) that controls a CPU or the like so as to realize the functions of the above-described embodiment. Then, the information handled by these devices is temporarily stored in a temporary storage device (for example, RAM) at the time of processing, and then stored in storage devices of various ROMs, HDDs, and SSDs, and read by the CPU as needed. , Correction and writing are performed.

  In the case of distribution in the market, the program can be stored and distributed in a portable recording medium, or can be transferred to a server computer connected via a network such as the Internet. In this case, of course, the storage device of the server computer is also included in the present invention.

Reference Signs List 1 abnormality notifying system 3 detection device 5 processing device 100 control unit 200 biological signal acquisition unit 300 biological information value calculation unit 350 sleep state determination unit 400 input unit 450 output unit 500 storage unit 510 biological information data 520 sleep state data 600 patient condition acquisition Part 650 Patient diary output part 700, 750 Patient state estimation part 710 Feature extraction part 720 Identification part 730 Identification dictionary 740 Patient state output part 800 Learning part 900 Alert output part 10 Bed 20 Mattress

Claims (3)

Biological signal acquisition means for acquiring a biological signal in the bed of the subject,
Biological information value calculation means for calculating a biological information value from the acquired biological signal;
Estimating means for estimating a state of the subject after an estimation time has elapsed based on the biological information value;
Notification means for notifying when the condition of the subject is determined to be abnormal by the estimation means;
State determination means for determining the current state of the subject based on the biological information value;
Equipped with
The informing device infers the state of the subject based on the result of learning from the current state of the subject determined by the state determining means.   The state determination means determines that tachypnea continues for a predetermined time or more, and that the respiration rate and the heart rate decrease by each change amount within the predetermined time, or the patient's departure judgment continues for a predetermined time or more in a patient who can not leave the bed The abnormality notification device according to claim 1, wherein the current state of the subject is determined by any of the following. On the computer
A biological signal acquisition function of acquiring a biological signal in the bed of the subject,
A biological information value calculating function of calculating a biological information value from the acquired biological signal;
An estimation function for estimating a state of the subject after an estimation time has elapsed based on the biological information value;
A notification function for notifying when the condition of the subject is determined to be abnormal by the estimation function;
A state determination function of determining the current state of the subject based on the biological information value;
A program to realize
The program for estimating the condition of the subject based on the result of learning from the condition of the current subject determined by the condition determining function.

Images