Pain detection device, pain detection method, and pain detection program

Abstract

To provide a pain detection device, a pain detection method, and a pain detection program that can detect pain with high accuracy.SOLUTION: A pain detection device includes: data reception means for receiving vital data and pain data of an object person; learning means for generating a pain detection model that has learned the vital data in a normal time; and detection means for detecting pain on the basis of the pain detection model and detecting vital data of the object person. The learning means corrects the pain detection model on the basis of a pain state that is received from the object person.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a pain detection device, a pain detection method, and a pain detection program. [Background technology]

[0002] When dealing with pain, it is desirable to select an appropriate measure depending on the intensity of the pain. Because pain is subjective, it is difficult to objectively assess the intensity of pain by interviewing the patient. Therefore, technology for assessing pain intensity using biological data collected from the patient is being developed. For example, Patent Document 1 discloses a technology for analyzing the pain level based on the patient's electroencephalogram. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2020-203121 Summary of the Invention [Problem to be solved by the invention]

[0004] The technology disclosed in Patent Document 1 generates a learning model for determining pain levels and estimates the pain level using the learning model. However, it does not confirm whether the estimated pain level matches the patient's physical sensation, and there is room for further improvement in the accuracy of pain detection.

[0005] In view of the above-mentioned problems, an object of the present disclosure is to provide a pain detection device, a pain detection method, and a pain detection program that are capable of detecting pain with high accuracy. [Means for solving the problem]

[0006] The pain detection device according to the present disclosure comprises: data receiving means for receiving vital data and pain data of the subject; a learning means for generating a pain detection model that has learned the vital data in a normal state; a detection means for detecting pain based on the pain detection model and vital data for detection of the subject; The learning means modifies the pain detection model based on the pain state received from the subject.

[0007] The pain detection method according to the present disclosure comprises: The computer Accept the subject's vital data and pain data, generating a pain detection model that has learned the vital data under normal conditions; Detecting pain based on the pain detection model and the subject's vital data for detection; The pain detection model is modified based on the pain condition received from the subject.

[0008] The pain detection program according to the present disclosure includes: A process of receiving vital data and pain data of a subject; A process of generating a pain detection model that has learned the vital data under normal conditions; A process of detecting pain based on the pain detection model and vital data for detection of the subject; modifying the pain detection model based on the pain state received from the subject; Let the computer run it. [Effects of the Invention]

[0009] The present disclosure makes it possible to provide a pain detection device, a pain detection method, and a pain detection program that are capable of detecting pain with high accuracy. [Brief explanation of the drawings]

[0010] [Figure 1] 1 is a block diagram showing a configuration of a pain detection device according to the present disclosure. [Figure 2]1 is a flowchart illustrating an example of the flow of a pain detection method according to the present disclosure. [Figure 3] 1 is a block diagram illustrating a configuration of a pain detection system according to the present disclosure. [Figure 4] FIG. 2 is a block diagram showing the configuration of a subject terminal. [Figure 5] 1 is a block diagram showing a configuration of a pain detection device according to the present disclosure. [Figure 6] 10 is a flowchart illustrating an example of a pain detection model generation process. [Figure 7] FIG. 10 is a diagram illustrating an example of data combination using time information. [Figure 8] FIG. 10 is a diagram illustrating an example of data combination using time information. [Figure 9] FIG. 10 is a diagram illustrating an example of data combination using time information. [Figure 10] 1 is a graph showing an example of changes in pain level and abnormality degree over time. [Figure 11] 10 is a flowchart illustrating an example of a pain detection process. [Figure 12] 10 is a flowchart illustrating an example of a time series prediction model generation process. [Figure 13] 10 is a flowchart illustrating an example of a pain prediction process. [Figure 14] FIG. 10 is a diagram showing the correlation of each parameter before and after the onset of pain. [Figure 15] FIG. 1 shows the results of a significant difference test for each parameter before and after the onset of pain. [Figure 16] FIG. 1 is a block diagram illustrating an example of a hardware configuration of a computer. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In each drawing, the same or corresponding elements are designated by the same reference numerals, and for clarity of explanation, duplicate explanations will be omitted as necessary.

[0012] <Embodiment 1> An example configuration of the pain detection device 100 will be described below with reference to FIG. 1. The pain detection device 100 includes a data receiving unit 110, a learning unit 120, and a detection unit 130. The pain detection device 100 is an information processing device that detects pain of a subject by using vital data of the subject. The subject may be, for example, a person experiencing chronic pain, such as a cancer patient, or a person experiencing various acute pains that occur in daily life. Examples of pain that occurs in daily life include headaches, abdominal pain, and menstrual pain.

[0013] The data receiving unit 110 receives vital data and pain data of the subject. The vital data is biological data of the subject and indicates the subject's vital activities. Examples of vital data include the number of steps, motion frequency (MF), body temperature, skin temperature, sweat rate, heart rate, pulse rate, respiratory rate, blood pressure, and oxygen saturation. The motion frequency is a numerical value indicating the frequency with which the subject's body moves per given period of time. The vital data may be data that can be measured using general-purpose equipment. For example, the vital data may be data measured using a commercially available wearable device, a smartphone, or the like. Time information is attached to the vital data. The time information attached to the vital data indicates the time at which the vital data was collected.

[0014] Pain data is data that numerically indicates the intensity of pain felt by a subject. The pain data may be recorded by the subject himself / herself, or may be recorded by a related party after interviewing the subject about the intensity of pain. The related party is a person related to the subject, such as the related party's family member or a medical professional who performs medical treatment on the related party. The pain data may also be recorded as a record of the pain intensity estimated by the related party based on the subject's appearance. Time information is attached to the pain data. The time information attached to the pain data indicates the time the pain data was collected. The pain data may be recorded at predetermined time intervals, at the same time as vital data, or when the pain intensity changes. The pain intensity, i.e., pain level, is generally evaluated using, but is not limited to, the Numerical Rating Scale (NRS) or the Visual Analogue Scale (VAS).

[0015] The learning unit 120 generates a pain detection model using vital data when the subject is not feeling pain as training data. Here, "the subject is not feeling pain" refers to the pain level in the pain data being below a predetermined level. Hereinafter, "when the subject is not feeling pain" may be referred to as "normal." The learning unit 120 links the pain data to the vital data using the time information attached to the vital data and the time information attached to the pain data. Next, the learning unit 120 extracts sections of the vital data where the pain level is below a predetermined level, and generates a pain detection model trained using the vital data of the extracted sections, i.e., the normal vital data, as training vital data. When there are multiple subjects, the learning unit 120 generates a pain detection model for each subject.

[0016] The detection unit 130 detects pain based on a pain detection model and detection vital data. The detection unit 130 detects pain using supervised anomaly detection, which is a type of machine learning method. Examples of supervised anomaly detection include, but are not limited to, k-nearest neighbor and one-class support vector machines. An example of pain detection by the detection unit 130 using the k-nearest neighbor method will be described below. The k-nearest neighbor method calculates the distance between new data, i.e., test data, and the training vital data, and performs classification by drawing a circle containing a predetermined number of k data points. Abnormalities are detected by utilizing the characteristic that a close distance is normal and a far distance is abnormal. Training is performed so that pain-free periods are considered normal and periods where pain occurs are detected as abnormal values.

[0017] The subject or a related person records the intensity of pain felt by the subject at the time the detection vital data was collected as pain data. The recorded pain data is accepted by the data accepting unit 110. If there is a discrepancy between the detection result by the detection unit 130 and the pain data accepted by the data accepting unit 110, the learning unit 120 corrects the pain detection model by re-learning it based on the pain data.

[0018] Next, an example of a pain detection method according to the present disclosure will be described with reference to FIG. 2. First, the data accepting unit 110 accepts vital data and pain data of the subject (step S101). Next, the learning unit 120 extracts normal vital data based on the vital data and pain data accepted in step S101, and generates a pain detection model trained on the normal vital data (step S102). Next, the detection unit 130 detects pain based on the detection vital data accepted by the data accepting unit 110 (step S103). Next, the learning unit 120 modifies the pain detection model based on the pain data at the time of acquiring the detection vital data accepted by the data accepting unit 110, i.e., the actual pain state (step S104).

[0019] In this way, the pain detection method using the pain detection device 100 can improve the accuracy of pain detection in real time by feeding back the intensity of pain actually felt by the subject. Therefore, highly accurate pain detection can be performed using vital data collected in real time.

[0020] The pain detection device 100 includes a processor, a memory, and a storage device (not shown). The storage device stores a computer program that implements the processing of the pain detection method according to this embodiment. The processor then loads the computer program from the storage device into the memory and executes the computer program. This allows the processor to implement the functions of the data receiving unit 110, the learning unit 120, and the detection unit 130.

[0021] Alternatively, each component of the pain detection device 100 may be realized by dedicated hardware. Furthermore, some or all of the components of each device may be realized by general-purpose or dedicated circuits, processors, etc., or a combination of these. These may be configured by a single chip, or by multiple chips connected via a bus. Some or all of the components of each device may be realized by a combination of the above-mentioned circuits, etc., and a program. Furthermore, a CPU (Central Processing Unit), GPU (Graphics Processing Unit), FPGA (Field-Programmable Gate Array), quantum processor (quantum computer control chip), etc., may be used as the processor.

[0022] Furthermore, when some or all of the components of the pain detection device 100 are realized by multiple information processing devices, circuits, etc., the multiple information processing devices, circuits, etc. may be centrally or decentralized. For example, the information processing devices, circuits, etc. may be realized as a client-server system, a cloud computing system, or the like, in a form in which each is connected via a communication network. Furthermore, the functions of the pain detection device 100 may be provided in a SaaS (Software as a Service) format.

[0023] <Embodiment 2> The pain detection device 400 shown in FIG. 3 is an example of the pain detection device 100 described above. An example configuration of a pain detection system 500 including the pain detection device 400 will be described with reference to FIG. 3. The pain detection system 500 is communicably connected to a related party terminal 800 and a subject terminal 900 via a network 300. The network 300 connects information and communication devices via wireless communication lines, and may be a network accessible only to some information devices, or a network accessible to an unspecified number of people, such as the Internet. The related party terminal 800 is a communication terminal that can be operated by the related party, such as a smartphone, tablet, or PC. The subject terminal 900 is a communication terminal that can be operated by the subject, such as a commercially available wearable device.

[0024] As shown in FIG. 3, the pain detection system 500 includes a pain detection device 400, a learning model database 600, and a subject database 700. The pain detection system 500 is an information processing system that detects and predicts pain in a subject. The learning model database 600 is a database in which various generated learning models are stored. The subject database 700 is a database in which data related to subjects is stored, and vital data, pain data, etc. are stored for each subject.

[0025] FIG. 4 is a block diagram showing an example configuration of the subject terminal 900. As shown in FIG. 4, the subject terminal 900 includes a vital data acquisition unit 920, a pain data acquisition unit 930, a data transmission unit 940, a notification reception unit 950, an exercise status data acquisition unit 960, a lifestyle habit data acquisition unit 970, and a location information data acquisition unit 980. The vital data acquisition unit 920 acquires vital data of the subject. If the subject terminal 900 is a wearable device, the vital data acquisition unit 920 collects the vital data while the subject is wearing the subject terminal 900. The pain data acquisition unit 930 acquires pain data by accepting input of a pain level by the subject. The data transmission unit 940 transmits various data acquired by the subject terminal 900 to an external device. The notification reception unit 950 receives the pain detection result and the pain prediction result transmitted by the pain detection system 500 and notifies the subject of the results.

[0026] The exercise status data acquisition unit 960 acquires exercise status data of the subject. The exercise status data is data that changes when the subject exercises, and examples include the number of steps, amount of exercise, and calories burned. The exercise status data is accompanied by time information. The time information accompanied by the exercise status data is information indicating the time when the exercise status data was collected. Generally, when the subject exercises, vital data such as heart rate and body temperature increase. Therefore, there is a risk that a change in vital data due to exercise may be erroneously detected as a change in vital data due to the occurrence of pain. Therefore, as will be described in detail later, the pain detection system 500 generates a pain detection model that takes into account the subject's exercise status.

[0027] The lifestyle data acquisition unit 970 acquires lifestyle data of the subject. The lifestyle data is data related to the subject's lifestyle, such as an exercise cycle acquired based on one week's worth of exercise status data. Specifically, the lifestyle data is information such as the subject's habit of walking to work or school at a specific time on weekdays. The location information data acquisition unit 980 acquires the subject's location information. The location information is measured using, for example, a Global Navigation Satellite System (GNSS), but is not limited to this. The lifestyle data acquisition unit 970 may acquire the subject's exercise cycle, etc., based on the location information acquired by the location information data acquisition unit 980.

[0028] Next, an example of the configuration of the pain detection device 400 will be described with reference to Fig. 5. The pain detection device 400 is an information processing device that performs pain detection processing, etc., and is, for example, a server device realized by a computer. The pain detection device 400 may be redundantly configured with multiple servers, and each functional block may be realized by multiple computers. The pain detection device 400 includes a memory 410, a communication unit 420, a storage unit 430, and a control unit 440.

[0029] The memory 410 is a storage area that temporarily stores the processing contents of the control unit 440, and is a volatile storage device such as a RAM (Random Access Memory). The communication unit 420 is an interface that communicates with the outside of the pain detection device 400. The storage unit 430 is a storage device that stores a program 431 and the like. The program 431 is a computer program that implements the pain detection processing according to the present disclosure.

[0030] The control unit 440 includes a data receiving unit 441, a learning unit 442, a prediction unit 443, a detection unit 444, and a notification unit 445. The control unit 440 is a control device that controls the operation of the pain detection device 400, and is, for example, a processor such as a CPU. The control unit 440 loads the program 431 from the storage unit 430 into the memory 410 and executes it. In this way, the control unit 440 realizes the functions of the data receiving unit 441, the learning unit 442, the prediction unit 443, the detection unit 444, and the notification unit 445.

[0031] The subject terminal 900 transmits various data such as vital data, pain data, and exercise status data to the pain detection device 400. When the data receiving unit 441 receives the various data, it accepts the various data and stores it in the subject database 700. The subject database 700 manages the stored various data individually for each subject.

[0032] Of the various data stored in the subject database 700, the vital data, pain data, and exercise status data are loaded into the learning unit 442. The learning unit 442 generates a pain detection model for each exercise status label based on the vital data, pain data, and exercise status data. FIG. 6 shows an example of the flow of the pain detection model generation process. In the pain detection model generation process, the learning unit 442 first loads the vital data, pain data, and exercise status data for the subject for which a model is to be generated from the subject database 700 as learning data (step S201).

[0033] Next, the learning unit 442 combines these pieces of learning data based on the time information attached to the vital data, pain data, and exercise status data (step S202). FIGS. 7 to 9 show examples of combining pain data and vital data. For example, as shown in FIG. 7, the learning data may be complemented by using the immediately preceding value as the value of one piece of learning data at the time when the data is missing. Alternatively, as shown in FIG. 8, the learning data may be complemented by taking a weighted average of the preceding and following values ​​as the value of one piece of learning data at the time when the data is missing. Alternatively, as shown in FIG. 9, the learning data may be extracted from only the data at the time when both pieces of learning data are present. Note that the method of combining data is not limited to these.

[0034] Returning to FIG. 6, the description continues. Next, the learning unit 442 assigns an exercise status label to the vital data based on the exercise status data (step S203). The exercise status label may be assigned by any method. If the subject's device 900 has a function for automatically determining the subject's exercise status, the exercise status determined using that function may be used as the exercise status label. Furthermore, the exercise status may be labeled by setting a threshold for each subject's exercise status data, such as calories burned, exercise volume, and number of steps. If lifestyle habit data indicates commuting to work or school on a specific day and time, the specific day and time may be considered to be an exercise state, and other times may be considered to be a resting state. Temperatures above a specific level based on body surface temperature or body temperature may be considered to be an exercise state, and temperatures below the specific level may be considered to be a resting state. Furthermore, the exercise status may be labeled by manually recording the start and end times of exercise. Furthermore, the exercise status label may be assigned based on the subject's location information data.

[0035] Next, the learning unit 442 extracts, from the vital data, a section in which the subject is not feeling pain as normal vital data (step S204). The upper part of FIG. 10 is a graph showing changes in pain data over time. In the section surrounded by the dotted line in the upper part of FIG. 10, the pain level is low and the subject is not feeling pain. Therefore, the learning unit 442 extracts this section as normal vital data.

[0036] Returning to FIG. 6, the explanation will be continued. Next, the learning unit 442 divides the normal vital data for each exercise condition label (step S205). Next, the learning unit 442 generates a pain detection model trained for each exercise condition label based on the divided normal vital data (step S206). The generated pain detection models are each stored in the learning model database 600 (step S207). In this way, the learning unit 442 generates a pain detection model for each exercise condition label.

[0037] The detection unit 444 detects pain based on the vital data and the pain detection model. FIG. 11 shows an example of the flow of the pain detection process. In the pain detection process, the detection unit 444 first acquires the pain detection model by loading it from the learning model database 600 (step S301). Next, the detection unit 444 acquires vital data for detection and exercise status data including the exercise status at the time corresponding to the vital data (step S302). The vital data for detection is vital data for which pain detection is desired, and may be acquired from the subject database 700. Furthermore, when pain detection is performed in real time, the vital data for detection may be acquired directly from the subject terminal 900.

[0038] Next, the detection unit 444 detects pain using a pain detection model for the vital data for detection (step S303). Specifically, the detection unit 444 attaches an exercise status label to the vital data for detection and performs pain detection using a pain detection model corresponding to the exercise status label. For example, the detection unit 444 may calculate an abnormality level indicating how much the vital data for detection differs from normal vital data, and determine that pain is occurring when the abnormality level is equal to or greater than a threshold. The lower part of FIG. 10 is a graph showing the abnormality level calculated based on the vital data. In the example shown in the lower part of FIG. 10, pain is detected using a threshold value of 0.4. In this way, the detection unit 444 detects pain based on the vital data.

[0039] Next, the notification unit 445 transmits the pain detection result to at least one of the related person terminal 800 and the subject terminal 900 (step S304). When the detection result is transmitted to the subject terminal 900, the notification result including a message requesting the subject to input the intensity of pain actually felt during the detection period may be displayed on the subject terminal 900. The subject operates the subject terminal 900 to input the intensity of pain actually felt. Note that the subject may input the intensity of pain only when the notified detection result differs from the intensity of pain actually felt. The subject terminal 900 transmits the input pain intensity to the pain detection device 400 as pain data at the time of detection. The pain data at the time of detection is received by the data receiving unit 441 as feedback data (step S305).

[0040] The learning unit 442 compares the detection result with the feedback data and determines whether the detected pain level matches the pain level actually felt by the subject (step S306). If the detected pain level does not match the pain level actually felt by the subject (step S306 No), the learning unit 442 re-learns the pain detection model based on the feedback data (step S307), and stores the re-learned pain detection model in the learning model database 600 (step S308). Since the pain detection model can be fed back in this way, the pain detection device 400 can detect pain with higher accuracy.

[0041] If the detected pain level matches the pain level actually felt by the subject (Yes in step S306), the subject selects a method of dealing with the pain. Specifically, for example, if the subject is a cancer patient who manages pain by taking rescue medication that requires a doctor's prescription, the subject may consult a medical professional who owns the relevant person terminal 800 about taking the rescue medication based on the pain detection result. If the detected pain level matches the pain level actually felt by the subject (Yes in step S306), the notification unit 445 may send a notification to the subject's terminal 900 including a message urging the subject to consult a medical professional about medication.

[0042] In this way, by evaluating the intensity of pain felt by a subject based on vital data, the intensity of pain felt by the subject can be objectively evaluated. Therefore, overdose due to chemical coping can be prevented. Chemical coping is a phenomenon in which a subject who may experience pain, such as cancer pain, takes rescue medication such as opioids due to anxiety even when no pain is occurring. Furthermore, this can prevent a subject from refraining from taking medication even when the pain is at a level that would warrant taking rescue medication.

[0043] The pain detection device 400 detects pain based on vital data that can be collected using a general-purpose mobile communication terminal. In other words, the pain detection device 400 does not require dedicated biometric data acquisition equipment that can only be used in medical institutions, etc. Therefore, the pain detection device 400 can detect pain in subjects who are not hospitalized but are living at home, etc. Furthermore, the pain detection device 400 can detect not only the above-mentioned cancer pain but also acute pain that occurs in daily life.

[0044] By notifying the related party terminal 800 of the detection results in step S304, the related party can understand the degree of pain of the subject, which may lead to pain treatment that is in line with the subject's lifestyle. When notifying the related party terminal 800, the notification may be sent together with location information data transmitted from the subject terminal 900. When the subject's pain is severe, it may be desirable for the related party to rush to the subject's side. However, when the pain is severe, it may be difficult for the subject to communicate by phone or the like. By notifying the subject's location information data together with the pain detection results, the related party can see the notified location information data and rush to the subject's side.

[0045] In addition to the pain detection process, the pain detection device 400 may also perform a process of predicting the occurrence of pain. In this case, the prediction unit 443 performs time-series prediction of the subject's vital data. Examples of time-series prediction include, but are not limited to, an autoregressive model, a Long Short Term Memory (LSTM), and a state space model. FIG. 12 shows an example of the flow of the time-series prediction model generation process. In the time-series prediction model generation process, the prediction unit 443 first acquires vital data, pain data, and exercise status data of the subject for which a model is to be generated from the subject database 700 by loading them (step S401).

[0046] Next, the prediction unit 443 combines these learning data based on the time information attached to the vital data, pain data, and exercise status data (step S402). Next, the prediction unit 443 assigns an exercise status label to the vital data based on the exercise status data (step S403). Next, the prediction unit 443 extracts a section in which the subject is not feeling pain from the vital data as normal vital data (step S404). Next, the prediction unit 443 divides the normal vital data for each exercise status label (step S405).

[0047] Next, the prediction unit 443 generates a time series prediction model trained for each exercise status label based on the divided normal vital data (step S406). The generated time series prediction models are stored in the learning model database 600 (step S407). In this way, the prediction unit 443 generates a time series prediction model for each exercise status label.

[0048] FIG. 13 is a flowchart showing an example of pain prediction processing. In the pain prediction processing, the prediction unit 443 first acquires vital data for prediction and exercise status data including the exercise status at the time corresponding to the vital data (step S501). The vital data for prediction is vital data for which time-series prediction is desired, and may be acquired from the subject database 700. Furthermore, when pain prediction is performed in real time, the vital data for prediction may be acquired directly from the subject terminal 900. Next, the prediction unit 443 combines the vital data and the exercise status data based on the time information assigned to each of the vital data and the exercise status data (step S502). That is, the prediction unit 443 assigns an exercise status label to the vital data for prediction. Next, the prediction unit 443 acquires a time-series prediction model corresponding to the exercise status label assigned to the vital data for prediction from the learning model database 600 by loading it (step S503). Next, the prediction unit 443 performs time-series prediction of the vital data based on the vital data for prediction and a time-series prediction model according to the exercise status label attached to the vital data for prediction (step S504).

[0049] Next, the detection unit 444 acquires a pain detection model corresponding to the exercise status label attached to the vital data for prediction from the learning model database 600 by loading it (step S505). Next, the detection unit 444 detects pain from the time-series predicted vital data using the pain detection model (step S506). Specifically, the detection unit 444 detects pain from the time-series predicted vital data using the pain detection model corresponding to the exercise status label attached to the time-series predicted vital data. Next, the notification unit 445 transmits the result of the pain prediction to at least one of the related party terminal 800 and the subject terminal 900 (step S507).

[0050] In this way, because the pain detection device 400 can predict the onset of pain, the subject and related parties can prepare pain relief measures in advance or take initial measures before the pain occurs. For example, when the subject is notified of the predicted onset of pain, the subject may consult with a medical professional about the painkiller to be used, the timing of use, the amount to be used, etc. before the pain occurs. Since pain relief measures can be prepared before the pain occurs, there is a possibility that a quick response can be made when pain occurs.

[0051] Next, with reference to Figures 14 and 15, we will explain the results of data analysis that support the feasibility of pain detection based on vital signs measured using a general-purpose device. To obtain the data shown in Figures 14 and 15, five subjects A to E simultaneously wore two commercially available wearable devices, and time-stamped pain level data was obtained. The two wearable devices were the Silmee W22 and the Fitbit Sense 2. First, a correlation analysis was performed between pain intensity and the data obtained from the wearable devices using data from the five minutes before pain onset (i.e., normal state data) and data during pain onset (i.e., pain state data). The correlation analysis results are shown in Figure 14. As shown in Figure 14, motion frequency, skin temperature, and pulse rate had a statistically significant positive correlation with pain intensity. This indicates that the values ​​of these data increase with pain intensity, demonstrating the feasibility of pain detection using methods such as calculating the degree of abnormality from normal vital signs data and time series analysis.

[0052] Next, we analyzed the data before and after the onset of pain, treating the five-minute data as a single unit. This analysis was conducted to verify whether the onset of pain could be detected promptly. Typical wearable devices often output various indicators on a minute-by-minute basis. Therefore, we set five-minute intervals to compare indicators that represent data changes, such as the mean, minimum, maximum, slope, and coefficient of variation. However, these intervals can be set to any value depending on the interval and characteristics of the data obtained. When acquiring the data shown in Figure 15, we labeled the five minutes before the onset of pain as normal data and the five minutes after the onset of pain as painful data, and performed statistical tests to determine whether there were any differences between the data before and after the onset of pain. Comparing the normal data with the data when moderate or severe pain was present, we found statistically significant differences in the number of steps, motion frequency, and pulse rate before and after the onset of pain, and each indicator increased when pain occurred.

[0053] In addition to the k-nearest neighbor method for individuals, pain detection using any time interval can also be performed using a time series analysis machine learning model such as LSTM for a group. This makes it possible to detect pain in subjects who are unable to report pain. Examples of subjects who are unable to report pain include, but are not limited to, infants and dementia patients. When performing pain detection on a subject who is unable to report pain, a family member or a medical professional or other related party may estimate the level of pain from the subject's appearance and input the estimated pain data into the subject's terminal 900. Furthermore, the subject is not limited to humans; pain detection can also be performed on animals using the same procedure.

[0054] Furthermore, if a device capable of acquiring heartbeat intervals and pulse wave intervals is used, it is possible to capture responses to pain using more data by calculating indices related to heart rate variability per time interval. In this case, the data is divided into arbitrary time units and a dataset is created from which summary statistics are calculated. This dataset is then labeled using an arbitrary method for rest, exercise, and pain. Using the created dataset as training data, a classification model is created and pain detection is performed by classifying newly used patient data.

[0055] In the above example, a pain detection model was created for each subject, but a pain detection model may also be generated by learning vital data from multiple subjects. Furthermore, multiple subjects may be grouped according to attributes such as gender, age, and medical history, and a pain detection model may be generated for each group. In this case, too, the algorithm used to train the pain detection model is not particularly limited as long as it is a supervised anomaly detection algorithm.

[0056] <Example of hardware configuration> Hereinafter, with reference to FIG. 16, a case where each functional configuration of the pain detection device according to the present disclosure is realized by a combination of hardware and software will be described.

[0057] The pain detection device of the present disclosure can achieve the above-described functions by a computer 11 including the hardware configuration shown in FIG. 16. The computer 11 may be a portable computer such as a smartphone or tablet terminal, or a stationary computer such as a PC. The computer 11 may be a dedicated computer designed to realize each device, or may be a general-purpose computer. The computer 11 can achieve the desired functions by installing a predetermined program.

[0058] The computer 11 has a bus 21, a processor 30, a memory 40, a storage device 50, an input / output interface 60 (an interface is also called an I / F (Interface)), and a network interface 70. The bus 21 is a data transmission path through which the processor 30, the memory 40, the storage device 50, the input / output interface 60, and the network interface 70 transmit and receive data to and from each other. However, the method of connecting the processor 30 and the like to each other is not limited to bus connection.

[0059] The processor 30 is a variety of processors such as a CPU, a GPU, an FPGA, etc. The memory 40 is a main storage device realized using a RAM (Random Access Memory) or the like.

[0060] The storage device 50 is an auxiliary storage device realized using a hard disk, an SSD, a memory card, a ROM (Read Only Memory), or the like. The storage device 50 stores programs for realizing desired functions. The processor 30 reads these programs into the memory 40 and executes them to realize the various functional components of each device.

[0061] The input / output interface 60 is an interface for connecting the computer 11 with input / output devices. For example, the input / output interface 60 is connected to an input device such as a keyboard and an output device such as a display device.

[0062] The network interface 70 is an interface for connecting the computer 11 to a network.

[0063] Although an example of a hardware configuration for the present disclosure has been described above, the above-described embodiment is not limited to this. Any processing in the present disclosure can also be realized by causing a processor to execute a computer program.

[0064] In the above examples, the program includes instructions (or software code) that, when loaded into a computer, cause the computer to perform one or more functions described in the embodiments. The program may be stored on a non-transitory computer-readable medium or a tangible storage medium. By way of example and not limitation, computer-readable medium or tangible storage medium includes random-access memory (RAM), read-only memory (ROM), flash memory, solid-state drive (SSD) or other memory technology, CD-ROM, digital versatile disc (DVD), Blu-ray disc or other optical disk storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage device. The program may also be transmitted on a transitory computer-readable medium or communication medium. By way of example and not limitation, transitory computer-readable medium or communication medium includes electrical, optical, acoustic, or other forms of propagated signals.

[0065] Although the present disclosure has been described above with reference to the embodiments, the present disclosure is not limited to the above-described embodiments. Various modifications that can be understood by those skilled in the art can be made to the configuration and details of the present disclosure within the scope of the present disclosure. Furthermore, each embodiment can be combined with other embodiments as appropriate.

[0066] Each drawing is merely an example for describing one or more embodiments. Each drawing may relate not only to one particular embodiment, but also to one or more other embodiments. As will be understood by those skilled in the art, various features or steps described with reference to any one drawing can be combined with features or steps shown in one or more other drawings to create, for example, an embodiment not explicitly shown or described. Not all features or steps shown in any one drawing are necessary to describe an exemplary embodiment, and some features or steps may be omitted. The order of steps described in any drawing may be changed as appropriate.

[0067] A part or all of the above-described embodiments can be described as, but not limited to, the following supplementary notes.

[0068] (Appendix A1) data receiving means for receiving vital data and pain data of the subject; a learning means for generating a pain detection model that has learned the vital data in a normal state; a detection means for detecting pain based on the pain detection model and vital data for detection of the subject; the learning means corrects the pain detection model based on the pain state received from the subject; Pain detection device.

[0069] (Appendix A2) Further, a prediction means for predicting time series of vital data of the subject is provided, the detection means predicts the occurrence of pain based on time-series predicted vital data; 10. The pain detection device of claim 1.

[0070] (Appendix A3) Further provided is a notification means for notifying at least one of the subject and a person related to the subject of the detected result. 10. The pain detection device of claim A1 or A2.

[0071] (Appendix A4) The notification means notifies the relevant person of the location information data of the target person together with the detection result. 1. A pain detection device as described in Appendix A3.

[0072] (Appendix A5) the data receiving means receives vital data acquired by a wearable device worn by the subject; A pain detection device according to any one of appendices A1 to A4.

[0073] (Appendix A6) the data receiving means receives exercise status data indicating an exercise status of the subject; the learning means assigns an exercise status label to the vital data based on the exercise status data, and creates a pain detection model for each exercise status label of the subject; the detection means detects pain based on the vital data for detection of the subject and the pain detection model corresponding to the exercise status label attached to the vital data for detection; A pain detection device according to any one of appendices A1 to A5.

[0074] (Appendix A7) the data receiving means receives lifestyle habit data of the subject, the learning means creates a pain detection model for each exercise status label of the subject estimated based on the lifestyle habit data. 10. A pain detection device as described in Appendix A6.

[0075] (Appendix A8) the data receiving means receives location information data of the subject; the learning means creates a pain detection model for each exercise status label of the subject estimated based on the position information data. 10. A pain detection device as described in Appendix A6.

[0076] (Appendix A9) the data receiving means receives vital data with time information attached thereto and pain data with time information attached thereto; the learning means combines the vital data and the pain data by comparing time information with each other, and extracts, based on the pain data, a period in which the subject does not complain of pain as normal vital data; A pain detection device according to any one of appendices A1 to A8.

[0077] (Appendix B1) The computer Accept the subject's vital data and pain data, generating a pain detection model that has learned the vital data under normal conditions; Detecting pain based on the pain detection model and the subject's vital data for detection; modifying the pain detection model based on the pain status received from the subject; Pain detection methods.

[0078] (Appendix C1) A process of receiving vital data and pain data of a subject; A process of generating a pain detection model that has learned the vital data under normal conditions; A process of detecting pain based on the pain detection model and vital data for detection of the subject; modifying the pain detection model based on the pain state received from the subject; A pain detection program run on a computer.

[0079] Some or all of the elements (e.g., configurations and functions) described in Appendices A2 to A10 that are dependent on Appendix A1 may also be dependent on Appendix B1 and Appendix C1 in the same dependency relationship as Appendix A2 to A10. Some or all of the elements described in any appendix may be applied to various hardware, software, recording means for recording software, systems, and methods. [Explanation of symbols]

[0080] 100 Pain detection device 110 Data Reception Department 120 Learning Department 130 Detection unit 300 Network 400 Pain detection device 410 memory 420 Communications Department 430 Storage section 431 Program 440 Control Unit 441 Data Reception Department 442 Learning Department 443 Prediction Department 444 Detection unit 445 Notification Department 500 Pain Detection System 600 Learning Model Database 700 Subject Database 800 Terminal for related parties 900 Target Devices 920 Vital Data Acquisition Unit 930 Pain Data Acquisition Unit 940 Data Transmission Unit 950 Notification Receiving Unit 960 Exercise status data acquisition unit 970 Lifestyle Data Acquisition Department 980 Location Information Data Acquisition Unit 11. Computer 21 Bus 30 processors 40 memory 50 Storage Devices 60 Input / Output Interface 70 Network Interfaces

Claims

1. data receiving means for receiving vital data and pain data of the subject; a learning means for generating a pain detection model that has learned the vital data in a normal state; a detection means for detecting pain based on the pain detection model and vital data for detection of the subject; the learning means corrects the pain detection model based on the pain state received from the subject; Pain detection device.

2. Further, a prediction means for predicting time series of vital data of the subject is provided, the detection means predicts the occurrence of pain based on time-series predicted vital data; The pain detection device of claim 1 .

3. Further provided is a notification means for notifying at least one of the subject and a person related to the subject of the detected result.

3. The pain detection device according to claim 1 or 2.

4. the data receiving means receives vital data acquired by a wearable device worn by the subject; 3. The pain detection device according to claim 1 or 2.

5. the data receiving means receives exercise status data indicating an exercise status of the subject; the learning means assigns an exercise status label to the vital data based on the exercise status data, and creates a pain detection model for each exercise status label of the subject; the detection means detects pain based on the vital data for detection of the subject and the pain detection model corresponding to the exercise status label attached to the vital data for detection; 3. The pain detection device according to claim 1 or 2.

6. the data receiving means receives lifestyle habit data of the subject, the learning means creates a pain detection model for each exercise status label of the subject estimated based on the lifestyle habit data. The pain detection device of claim 5 .

7. the data receiving means receives location information data of the subject; the learning means creates a pain detection model for each exercise status label of the subject estimated based on the position information data. The pain detection device of claim 5 .

8. the data receiving means receives vital data with time information attached thereto and pain data with time information attached thereto; the learning means combines the vital data and the pain data by comparing time information with each other, and extracts, based on the pain data, a period in which the subject does not complain of pain as normal vital data; 3. The pain detection device according to claim 1 or 2.

9. The computer Accept the subject's vital data and pain data, generating a pain detection model that has learned the vital data under normal conditions; Detecting pain based on the pain detection model and the subject's vital data for detection; modifying the pain detection model based on the pain status received from the subject; Pain detection methods.

10. A process of receiving vital data and pain data of a subject; A process of generating a pain detection model that has learned the vital data under normal conditions; A process of detecting pain based on the pain detection model and vital data for detection of the subject; modifying the pain detection model based on the pain state received from the subject; A pain detection program run on a computer.

Images