Program, information processing method, information processing device and self-removal prevention system

Abstract

To provide a program that outputs symptom information regarding signs of self-removal of intravenous drip. [Solution] The program acquires time-series data regarding the movement of a second hand, different from the first hand being punctured, and outputs symptom information by inputting the acquired time-series data into a learning model that outputs symptom information regarding signs of self-removal of an IV drip when time-series data is input. Preferably, the program acquires video data of the puncture site and the movement of the second hand captured by a camera, and inputs the video data into the learning model to output the symptom information.

Description

[Technical Field]

[0001] The present disclosure relates to a program, an information processing method, an information processing device, and a self-removal prevention system. [Background technology]

[0002] Patent Document 1 discloses a self-removal monitoring system that automatically recognizes actions related to self-removal and notifies a monitor promptly and accurately. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6583953 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the invention of Patent Document 1 has a problem in that it is not possible to output symptom information regarding symptoms of self-removal of an intravenous drip.

[0005] In one aspect, an object of the present invention is to provide a program or the like that outputs symptom information regarding symptoms of self-withdrawal of an intravenous drip. [Means for solving the problem]

[0006] In the present invention, (1) a program acquires time series data regarding the movement of a second hand, which is different from the first hand being punctured, and causes a computer to execute a process of inputting the acquired time series data into a learning model that outputs symptom information regarding signs of self-removal of the IV drip when the time series data is input, thereby outputting symptom information. Here, in the embodiment of the present invention, (2) It is preferable that the program in (1) above acquires video data of the puncture site and the movement of the second hand captured by a camera, and outputs the symptom information by inputting the video data into the learning model. (3) It is preferable that the program of (1) or (2) acquires the video data that does not include a facial image of the patient, and outputs the symptom information by inputting the acquired video data into the learning model. (4) It is preferable that the program described in any of (1) to (3) above acquires time series data regarding the movement of the second hand obtained by an inertial sensor attached to the second hand, and outputs the symptom information by inputting the time series data into the learning model. (5) It is preferable that the program described in any of (1) to (4) above acquires first time series data regarding the movement of the first hand obtained by a first inertial sensor attached to the first hand, acquires second time series data regarding the movement of the second hand obtained by a second inertial sensor attached to the second hand, and outputs the symptom information by inputting the first time series data and the second time series data into the learning model. (6) In the program according to any one of (1) to (5) above, it is preferable that the symptom information indicates a probability of self-removal, and an alert is output when the probability is equal to or greater than a predetermined threshold. (7) In the program described in any one of (1) to (6) above, it is preferable that the symptom information indicates a probability of self-removal, and an alert is output when the probability is equal to or greater than a predetermined threshold and the distance between the first hand and the second hand is within a predetermined range. (8) It is preferable that the program described in any one of (1) to (7) above outputs a command to give the patient a stimulus to prevent self-removal when the probability is equal to or greater than a predetermined threshold. (9) It is preferable that the program according to any one of (1) to (8) above outputs a second alert when self-removal is detected. (10) In the program according to any one of (1) to (9) above, it is preferable that the predetermined threshold value or the content of the alert be changed depending on the condition of the patient. An information processing method (11) according to one aspect of the present disclosure acquires time series data regarding the movement of a second hand that is different from the first hand being punctured, and outputs symptom information by inputting the acquired time series data into a learning model that outputs symptom information regarding signs of self-removal of an IV drip when time series data is input. (12) An information processing device according to one aspect of the present disclosure includes a control unit that acquires time series data regarding the movement of a second hand that is different from the first hand being punctured, and executes a process of outputting symptom information by inputting the acquired time series data into a learning model that outputs symptom information regarding signs of self-removal of an IV drip when the time series data is input. (13) A self-removal prevention system according to one embodiment of the present disclosure includes a sensor that acquires time series data regarding the movement of a second hand that is different from the first hand being punctured, and an information processing device that inputs the acquired time series data into a learning model that outputs symptom information regarding signs of self-removal of an IV drip when the time series data is input, and outputs the symptom information. [Effects of the Invention]

[0007] In one aspect, symptom information regarding symptoms of self-removal of an IV drip can be output. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is an explanatory diagram showing an overview of a self-removal prevention system. [Figure 2] FIG. 2 is a block diagram showing an example of the configuration of a tablet. [Figure 3] FIG. 2 is a block diagram illustrating an example of the configuration of a server. [Figure 4] FIG. 1 is a block diagram illustrating an example of the configuration of a smartphone. [Figure 5] FIG. 2 is an explanatory diagram showing an example of a record layout of a patient DB. [Figure 6] FIG. 1 is an explanatory diagram of a learning model. [Figure 7] FIG. 10 is an explanatory diagram showing an example of a record layout of a training DB. [Figure 8A] FIG. 10 is an explanatory diagram showing an example of an alert screen for a patient. [Figure 8B] FIG. 10 is an explanatory diagram showing an example of an alert screen for medical personnel. [Figure 9] 10 is a flowchart showing a procedure for generating a learning model. [Figure 10] 10 is a flowchart illustrating an example of a processing procedure of the self-removal prevention system. [Figure 11] FIG. 10 is an explanatory diagram showing an overview of a self-removal prevention system according to a second embodiment. [Figure 12] FIG. 2 is a block diagram showing an example of the configuration of a tablet and an inertial sensor. [Figure 13] FIG. 10 is an explanatory diagram of a learning model according to the second embodiment. [Figure 14] FIG. 10 is an explanatory diagram showing an example of a record layout of a training DB according to the second embodiment. [Figure 15] 10 is a flowchart showing a procedure for generating a learning model according to the second embodiment. [Figure 16] 10 is a flowchart showing an example of a processing procedure of the self-removal prevention system according to the second embodiment. [Figure 17] FIG. 10 is an explanatory diagram showing an overview of a self-removal prevention system according to a third embodiment. [Figure 18] FIG. 2 is a block diagram showing an example of the configuration of a tablet, a first inertial sensor, and a second inertial sensor. [Figure 19] FIG. 10 is an explanatory diagram of a learning model according to the third embodiment. [Figure 20] FIG. 11 is an explanatory diagram showing an example of a record layout of a training DB according to the third embodiment. [Figure 21] 10 is a flowchart showing a procedure for generating a learning model according to the third embodiment. [Figure 22] 11 is a flowchart showing an example of a processing procedure of the self-removal prevention system according to the third embodiment. [Figure 23]FIG. 11 is an explanatory diagram showing an example of a record layout of a patient DB according to the fourth embodiment. [Figure 24A] FIG. 10 is an explanatory diagram showing an example of a screen displaying a changed patient alert. [Figure 24B] FIG. 10 is an explanatory diagram showing an example of a screen displaying a modified alert for medical personnel. [Figure 25] 10 is a flowchart showing an example of a processing procedure of the self-removal prevention system according to the fourth embodiment. [Figure 26] 10 is a flowchart showing an example of a processing procedure of the self-removal prevention system according to the fourth embodiment. [Figure 27] FIG. 10 is a block diagram showing an example of the configuration of a tablet and an inertial sensor according to a fifth embodiment. [Figure 28] 13 is a flowchart showing an example of a processing procedure of the self-removal prevention system according to the fifth embodiment. [Figure 29] 13 is a flowchart showing an example of a processing procedure of the self-removal prevention system according to the sixth embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0009] (Embodiment 1) FIG. 1 is an explanatory diagram showing an overview of a self-removal prevention system. The self-removal prevention system 10 acquires time-series data related to the movement of a patient's second hand, which is different from the first hand that is being punctured, and outputs symptom information related to symptoms of self-removal of an IV drip by inputting the acquired time-series data into a learning model. The time-series data includes video data captured by a camera or time-series data obtained by an inertial sensor. In the first embodiment, an example will be described in which the self-removal prevention system 10 outputs symptom information related to symptoms of self-removal of an IV drip by inputting video data captured by a camera (see FIG. 2) into a learning model.

[0010] The self-removal prevention system 10 includes an information processing device 20, an information processing device 30, and an information processing device 40. The information processing device 20, the information processing device 30, and the information processing device 40 are connected via a network N. The information processing device 20 is, for example, a server device, a personal computer, or a general-purpose tablet PC (personal computer). In this embodiment, the information processing device 20 will be described as a tablet 20. The tablet 20 captures an image of the puncture site (left wrist in FIG. 1) of a patient lying on a bed or the like, and the movement of the patient's second hand (right hand in FIG. 1) that is different from the first hand (left hand in FIG. 1) that is being punctured. The patient is, for example, an elderly person living in each room in a facility such as a nursing home or a medical facility. A tablet 20 is provided for each patient. In FIG. 1, two tablets 20 are provided for patient A and patient B, respectively. The tablets 20 and the patients are associated with each other by the patient's identification information (hereinafter referred to as patient ID). The tablet 20 is placed on top of an IV stand. In addition to the tablet 20, an infusion pump is also placed on the IV stand. The infusion pump is placed below the tablet 20. The IV stand is placed, for example, beside the patient's bed.

[0011] The information processing device 30 is an information processing device that processes, stores, and transmits / receives various types of information. The information processing device 30 is, for example, a server device, a personal computer, or a general-purpose tablet PC (personal computer). The information processing device 30 may be a cloud server device that provides functions included in the information processing device 30 as cloud services. In this embodiment, the information processing device 30 will be described as a server 30.

[0012] The information processing device 40 is an information processing device used by a medical professional. The information processing device 40 is, for example, a server device, a smartphone, a tablet, a personal computer, or a general-purpose tablet PC (personal computer). In this embodiment, the information processing device 40 will be described as a smartphone 40.

[0013] FIG. 2 is a block diagram showing an example configuration of a tablet. The tablet 20 includes a control unit 21, a storage unit 22, a communication unit 23, a display unit 24, a speaker 25, an input unit 26, and a camera 27. The above-mentioned units are interconnected via a bus. The control unit 21 is configured using one or more processors, such as a central processing unit (CPU), a microprocessing unit (MPU), or a graphics processing unit (GPU). The storage unit 22 includes a random access memory (RAM) or a read-only memory (ROM). The storage unit 22 stores various data required for a control program 22P (a program product) executed by the control unit 21. The storage unit 22 also temporarily stores data generated when the control program 22P is executed. The control unit 21 appropriately executes the control program 22P stored in the storage unit 22 to perform various information processing and control processing related to the tablet 20. The communication unit 23 is a communication module that transmits and receives information to and from the server 30 via the network N.

[0014] The display unit 24 includes, for example, a liquid crystal display or an organic EL (Electro Luminescence) display. The display unit 24 displays an alert to prevent the patient from removing the device themselves. The speaker 25 outputs the alert or the like as audio to prevent the patient from removing the device themselves. The speaker 25 may be externally connected to the tablet 20 via a wired or wireless connection. The input unit 26 is, for example, a touch sensor, and is stacked with the display unit 24 to form a touch panel. The input unit 26 may also be a keyboard or a mouse. The camera 27 has an imaging element such as a CCD (Charge Coupled Device) image sensor and a lens, and acquires image data related to the patient's puncture site and the movement of the second hand. The camera 27 sequentially acquires one image data (one frame). The camera 27 acquires video data at, for example, 60 frames, 30 frames, or 15 frames per second.

[0015] FIG. 3 is a block diagram showing an example of the configuration of a server. The server 30 includes a control unit 31, a storage unit 32, a communication unit 33, a mass storage unit 34, and a reading unit 35. The above-mentioned units are connected to each other via a bus. The control unit 31 is configured using a processor such as a CPU, an MPU, or a GPU. The storage unit 32 includes a RAM or a ROM. The storage unit 32 stores various data and the like required for a control program 32P (program product) executed by the control unit 31. The storage unit 32 also temporarily stores data and the like generated when executing the control program 32P. The control unit 31 executes the control program 32P stored in the storage unit 32 as needed to perform various information processing and control processing related to the server 30.

[0016] The communication unit 33 is a communication module that transmits and receives information between the tablet 20 and the smartphone 40 via the network N. The mass storage unit 34 includes RAM, ROM, or the like. The mass storage unit 34 stores a patient DB 341, a learning model 342, and a training DB 343. The patient DB 341, the learning model 342, and the training DB 343 will be described later. In this embodiment, the storage unit 32 and the mass storage unit 34 may be configured as an integrated storage device. The mass storage unit 34 may be configured by multiple storage devices. The mass storage unit 34 may be an external storage device connected to the server 30.

[0017] The reading unit 35 reads information stored in the portable storage medium 1a. The portable storage medium 1a is, for example, a CD (Compact Disc), a DVD (Digital Versatile Disc), a USB (Universal Serial Bus) memory, or an SD (Secure Digital). The reading unit 35 reads the control program 32P from the portable storage medium 1a. The control unit 31 stores the read control program 32P in the storage unit 32. The control unit 31 may download the control program 32P from another computer via the network N. In this case, the control unit 31 stores the downloaded control program 32P in the storage unit 32. The control unit 31 may store the read control program 32P in the mass storage unit 34.

[0018] In this embodiment, the server 30 may be configured with multiple servers. The server 30 may be a virtual machine virtually constructed by software within a single device. The server 30 may be a local server installed within the facility where the server 30 is located. The server 30 may be a cloud server connected for communication via a network N. Furthermore, the control program 32P may be executed on a single server, or may be distributed and executed across multiple servers interconnected via the network N.

[0019] FIG. 4 is a block diagram showing an example configuration of a smartphone. The smartphone 40 includes a control unit 41, a storage unit 42, a communication unit 43, and a display unit 44. The above-mentioned units are interconnected via a bus. The control unit 41 is configured using a processor such as a CPU, an MPU, or a GPU. The storage unit 42 includes a RAM or a ROM. The storage unit 42 stores various data required for a control program 42P (a program product) executed by the control unit 41. The storage unit 42 temporarily stores data generated when the control program 42P is executed. The control unit 41 executes the control program 42P stored in the storage unit 42 as needed to perform information processing and control processing related to the smartphone 40. The communication unit 43 is a communication module that transmits and receives information to and from the server 30 via the network N. The display unit 44 is a liquid crystal display, an organic electroluminescence (EL) display, or the like. The display unit 44 displays an alert to prevent the patient from removing the device themselves.

[0020] FIG. 5 is an explanatory diagram showing an example of a record layout of a patient DB. The patient DB 341 stores patient information. The patient DB 341 includes a patient ID column, a name column, and a room number column. The patient ID column stores a patient ID for identifying the patient. The name column stores the name of the patient. The room number column stores the room number in which the patient is located. In the record in FIG. 5 with patient ID K001, the name "A" and room number "Room 301" are stored.

[0021] FIG. 6 is an explanatory diagram of a learning model. Learning model 342 receives as input time-series data relating to the movement of a second hand, which is different from the first hand being punctured, and outputs symptom information relating to symptoms of self-removal of an IV drip based on the received time-series data. In embodiment 1, learning model 342 receives as input video data capturing the patient's puncture site (left wrist) and the movement of a second hand (right hand), which is different from the first hand (left hand) being punctured, as shown in FIG. 6, and outputs symptom information relating to symptoms of self-removal of an IV drip based on the received video data. The symptom information relating to symptoms of self-removal of an IV drip is the probability of self-removal, and is output as a value between 0 and 1.

[0022] The learning model 342 may use, as input data, time-series data in which the positions of the patient's joints are estimated, or time-series data in which the direction of the patient's face is estimated. When the control unit 31 uses, as input data, time-series data in which the positions of the patient's joints are estimated, the control unit 31 acquires, for example, image analysis means such as OpenPose, time-series data in which the positions of the patient's joints are estimated from video data (time-series image data) capturing images of the movement of the patient's second hand (right hand). When the control unit 31 uses, as input data, time-series data in which the direction of the patient's face is estimated, the control unit 31 acquires, for example, image analysis means such as OpenPose, time-series data in which the direction of the patient's face is estimated from video data capturing images of the entire patient. Furthermore, the learning model 342 may use, as input data, time-series data in which the positions of the patient's joints are estimated, and time-series data in which the direction of the patient's face is estimated.

[0023] The learning model 342 is configured using, for example, a Transformer. Note that the learning model 342 may be configured using other algorithms such as a Convolution Neural Network (CNN), a Recurrent Neural Network (RNN), or a Long Short Term Memory (LSTN), or may be configured by combining multiple algorithms. The learning model 342 may also be a language model such as a Generative Pre-trained Transformer (GPT). In this case, the learning model 342 receives video data as input and can output messages such as "Apparently the patient is about to remove the device themselves" or "Rest."

[0024] The learning model 342 has an input layer to which video data is input, an intermediate layer that extracts features from the input video data, and an output layer that outputs the probability of self-removal based on the calculation results of the intermediate layer. The input layer has input nodes to which video data (time-series image data) is sequentially input. The intermediate layer calculates output values ​​based on the video data input via the input layer using various functions, thresholds, etc. The output layer has output nodes that output the probability of self-removal. The output value from the output node is, for example, a value between 0 and 1. With the above-mentioned configuration, the learning model 342 outputs the probability of self-removal when video data is input.

[0025] In FIG. 6, when the control unit 31 inputs video data to the learning model 342, the learning model 342 outputs a probability of "0.8" of self-removal. If the probability of self-removal is equal to or greater than a predetermined threshold (e.g., 0.75), the control unit 31 acquires "self-removal will occur" as the estimated result. If the probability of self-removal is less than the predetermined threshold (e.g., 0.75), the control unit 31 acquires "self-removal will not occur" as the estimated result. In FIG. 6, the probability of self-removal of "0.8" is equal to or greater than the predetermined threshold, so the control unit 31 acquires "self-removal will occur" as the estimated result.

[0026] In FIG. 6, the probability of self-removal has been described as an example of symptom information regarding symptoms of self-removal of an IV drip, but this is not limiting. The learning model 342 may output the probability of self-removal (e.g., 0.7) and the probability of not self-removal (e.g., 0.3) as examples of symptom information regarding symptoms of self-removal of an IV drip. In this case, the sum of the probabilities output from the output layer is 1.0. The control unit 31 acquires the highest probability value from the multiple probability values ​​output from the output layer as the inferred result. In the above example, the control unit 11 acquires "self-removal" as the inferred result.

[0027] FIG. 7 is an explanatory diagram showing an example of a record layout of the training DB. The training DB 343 stores training data used to generate the learning model 342. The training DB 343 includes a video data sequence and a self-removal sequence. The video data sequence stores multiple video data, such as video data capturing signs of self-removal and video data not capturing signs of self-removal. The video data capturing signs of self-removal includes, for example, video data capturing the patient frequently looking at the puncture site, video data capturing the patient frequently touching the fixation portion of the puncture site, and video data capturing the patient's intense body movements. The video data capturing signs of self-removal includes, for example, video data capturing the patient frequently touching the IV tube, video data capturing the patient pulling the IV tube, video data capturing the patient peeling off a dressing, and video data capturing the patient frequently scratching the puncture site.

[0028] Examples of video data that do not capture signs of self-removal include video data that captures the patient not looking at the puncture site at all, video data that captures the patient not touching the fixation part at the puncture site at all, and video data that captures the patient's calm body movements. The self-removal column stores information indicating the presence or absence of signs of self-removal ("1": self-removal / "0": not self-removal). In the example of FIG. 7, video data that captures signs of self-removal and information indicating the presence or absence of signs of self-removal ("1": self-removal) are stored in association with each other. Also, in the example of FIG. 7, video data that does not capture signs of self-removal and information indicating the presence or absence of self-removal by the patient ("0": not self-removal) are stored in association with each other.

[0029] The training data may be time-series data estimating the positions of the patient's joints. In this case, the training DB 343 stores the time-series data estimating the positions of the patient's joints in association with information indicating whether the patient has performed self-removal. The training data may also be time-series data estimating the direction of the patient's face. In this case, the training DB 343 stores the time-series data estimating the direction of the patient's face in association with information indicating whether the patient has performed self-removal.

[0030] A method for generating the learning model 342 will be described. The control unit 31 reads out from the training DB 343 a plurality of training data in which video data is associated with information indicating the presence or absence of signs of self-removal. When video data for training is input to the learning model 342, the control unit 31 trains the learning model 342 so that it outputs the correct value (0 or 1) of the training data. The control unit 31 optimizes parameters used in the calculation process in the intermediate layer so that the output value from the output layer of the learning model 342 approaches the correct value of the training data. The parameters are, for example, node weights (coupling coefficients) or coefficients of activation functions used in each node. The parameter optimization method is, for example, backpropagation or steepest descent. The control unit 31 stores the learning model 342 generated by the above process in the mass storage unit 34.

[0031] The generation of the learning model 342 may be performed by an external computer or the like (not shown). In this case, the generated learning model 342 is deployed to the server 30 from the external computer or the like via the network N. In addition, although an example in which the learning model 342 is deployed to the server 30 will be described in this embodiment, this is not limiting. The learning model 342 may also be deployed to the tablet 20 or a computer or the like (not shown).

[0032] The processing of the first embodiment will be described. The control unit 21 uses the camera 27 to acquire video data capturing the patient's puncture site and the movement of the second hand. At this time, the medical staff may change the position or angle of the tablet 20 so that the image does not include the patient's facial image, in consideration of the patient's privacy. In this case, the control unit 21 can acquire video data that does not include the patient's facial image. In subsequent processing, the control unit 21 may use either video data that includes the patient's facial image or video data that does not include the patient's facial image. The control unit 21 divides the acquired video data into video data of a predetermined time (hereinafter referred to as 30 seconds). The control unit 21 divides the video data so that the previously divided video data and the next video data to be divided include overlapping portions. Specifically, when dividing the acquired video data into 30-second video data, the control unit 21 divides the video data so that the previously divided video data and the next video data to be divided include 15 seconds of video data that overlap. When dividing the acquired video data, the control unit 21 may divide the video data so that the previously divided video data and the next divided video data do not include overlapping portions. The control unit 21 sequentially transmits the divided video data (30 seconds of video data) and the patient ID to the server 30.

[0033] The control unit 31 receives the video data and the patient ID transmitted from the tablet 20. The control unit 31 reads the patient's name and room number associated with the patient ID from the patient DB 341. The control unit 31 inputs the received video data into the learning model 342. The control unit 31 acquires the probability of self-removal output by the learning model 342. The control unit 31 determines whether the acquired probability is equal to or greater than a predetermined threshold. If the control unit 31 determines that the acquired probability is not equal to or greater than the predetermined threshold, it receives the video data transmitted from the tablet 20 again. If the control unit 31 determines that the acquired probability is equal to or greater than the predetermined threshold, it generates an alert to prevent the patient from self-removing the device using the patient's name and room number. The alert to prevent the patient from self-removing the device includes an alert to notify the patient of self-removal (hereinafter referred to as a patient alert) and an alert to notify a medical professional of self-removal (hereinafter referred to as a medical professional alert). The control unit 31 transmits the patient alert to the tablet 20. The control unit 31 also transmits an alert to the smartphone 40 for medical personnel.

[0034] The control unit 21 receives a patient alert sent from the server 30. Upon receiving the patient alert from the server 30, the control unit 21 immediately displays the received patient alert on the display unit 24. FIG. 8A is an explanatory diagram showing an example of a patient alert screen. The screen d1 in FIG. 8A includes a patient information field d11 and a patient message field d12. The patient information field d11 displays the patient's room number and name. The patient information field d11 in FIG. 8A displays "Room 301, Mr. A." The patient message field d12 displays a message for the patient. The patient message field d12 in FIG. 8A displays "Signs of self-removal have been observed. Do not remove the IV drip." The control unit 21 outputs a warning sound (e.g., a beep) and the content of the patient message field d12 as voice from the speaker 25 in parallel with the display of the screen d1.

[0035] The control unit 41 receives an alert for a medical worker transmitted from the server 30. Upon receiving the alert for a medical worker from the server 30, the control unit 41 immediately displays the received alert for a medical worker on the display unit 44. FIG. 8B is an explanatory diagram showing an example of a screen for an alert for a medical worker. The screen d2 in FIG. 8B includes a message field for a medical worker d21. The message field for a medical worker d21 displays a message for the medical worker. The message field for a medical worker d21 in FIG. 8B displays, "Mr. A in room 301 is attempting to remove his / her IV drip himself / herself. Please head to the scene." The control unit 41 may output the content of the message field for a medical worker d21 as audio from a speaker (not shown) of the smartphone 40 in parallel with the display of the screen d2.

[0036] The control unit 41 may display video data capturing signs of self-removal on the screen d2 in addition to the message field d21 for the medical staff. The processing in this case will be described below. When the control unit 31 determines that the acquired probability is equal to or greater than a predetermined threshold, it generates an alert for the medical staff using the patient's name and room number. Furthermore, when the control unit 31 determines that the acquired probability is equal to or greater than a predetermined threshold, it acquires the video data capturing signs of self-removal that were input to the learning model 342. The control unit 31 transmits the alert for the medical staff and the video data capturing signs of self-removal to the smartphone 40. The control unit 41 receives the alert for the medical staff and the video data capturing signs of self-removal transmitted from the server 30. The control unit 21 displays the received alert for the medical staff and the video data capturing signs of self-removal on the display unit 44. Instead of the video data capturing signs of self-removal, the control unit 21 may display a still image or a thumbnail of the video data capturing signs of self-removal on the display unit 44.

[0037] 9 is a flowchart showing the procedure for generating a learning model. The control unit 31 reads out from the training DB 343 a plurality of training data in which video data is associated with information indicating the presence or absence of signs of self-removal (step S101). The control unit 31 uses the read out training data to generate a learning model 342 that takes video data as input and outputs the probability of self-removal (step S102). The control unit 31 stores the generated learning model 342 in the mass storage unit 34 (step S103).

[0038] 10 is a flowchart showing an example of a processing procedure of the self-removal prevention system. The control unit 31 determines whether or not the video data and the patient ID transmitted from the tablet 20 have been received (step S201). If the control unit 31 determines that the video data and the patient ID transmitted from the tablet 20 have not been received (step S201: NO), the control unit 31 waits until the video data and the patient ID are received. If the control unit 31 determines that the video data and the patient ID transmitted from the tablet 20 have been received (step S201: YES), the control unit 31 reads the name and room number of the patient associated with the patient ID from the patient DB 341 (step S202). The control unit 31 inputs the received video data into the learning model 342 (step S203). The control unit 31 acquires the probability of self-removal output by the learning model 342 (step S204). The control unit 31 determines whether or not the acquired probability is equal to or greater than a predetermined threshold (step S205). If the control unit 31 determines that the acquired probability is not equal to or greater than the predetermined threshold (step S205: NO), the control unit 31 returns the process to step S201. If the control unit 31 determines that the acquired probability is equal to or greater than the predetermined threshold (step S205: YES), the control unit 31 generates an alert for the patient and an alert for the medical staff using the patient's name and room number (step S206). The control unit 31 transmits the alert for the patient to the tablet 20 (step S207). In parallel with the process of step S207, the control unit 31 transmits an alert for the medical staff to the smartphone 40 (step S208).

[0039] The control unit 21 receives an alert for a patient transmitted from the server 30 (step S301). Upon receiving the alert for a patient from the server 30, the control unit 21 immediately displays the received alert for a patient on the display unit 24 (step S302). The control unit 41 receives an alert for a medical professional transmitted from the server 30 (step S401). Upon receiving the alert for a medical professional from the server 30, the control unit 41 immediately displays the received alert for a medical professional on the display unit 44 (step S402).

[0040] According to embodiment 1, the self-removal prevention system acquires time series data regarding the movement of a second hand, which is different from the first hand being punctured, and can output symptom information by inputting the acquired time series data into a learning model that outputs symptom information regarding symptoms of self-removal of the IV drip when time series data is input.

[0041] According to the first embodiment, the self-removal prevention system can acquire video data of the puncture site and the second hand movements captured by a camera, and can output symptom information by inputting the video data into the learning model.

[0042] According to the first embodiment, the self-removal prevention system can output an alert when the probability of self-removal is equal to or greater than a predetermined threshold.

[0043] According to the first embodiment, the self-removal prevention system can output the probability of self-removal while taking into consideration the privacy of the patient by using video data that does not include an image of the patient's face.

[0044] (Embodiment 2) In the second embodiment, an example is described in which a self-removal prevention system outputs the probability of self-removal by inputting time series data on the movement of the second hand obtained by an inertial sensor attached to the second hand into a learning model.

[0045] FIG. 11 is an explanatory diagram showing an overview of a self-removal prevention system according to a second embodiment. The self-removal prevention system 10 includes an inertial sensor 50. The inertial sensor 50 is attached to the wrist of a patient's second hand (e.g., the right hand) by a strap or the like. Alternatively, a wearable device (e.g., an Apple watch (registered trademark)) incorporating the inertial sensor 50 may be used instead of the inertial sensor 50. The inertial sensor 50 includes an acceleration sensor, a gyro sensor, and the like, and acquires time-series data related to the movement of the patient's second hand (the right hand in FIG. 11). The time-series data related to the movement of the patient's second hand is, for example, time-series data of triaxial acceleration related to the movement of the patient's second hand (hereinafter referred to as triaxial acceleration data) and time-series data of triaxial angular velocity related to the movement of the patient's second hand (hereinafter referred to as triaxial angular velocity data). The inertial sensor 50 and the tablet 20 are connected to each other so as to be able to communicate with each other via short-range wireless communication such as Bluetooth (registered trademark). The tablet 20, the inertial sensor 50, and the patient are associated with each other by the patient ID.

[0046] In the second embodiment, an example in which the inertial sensor 50 includes an acceleration sensor and a gyro sensor is described, but this is not limiting. The inertial sensor 50 may include at least one of an acceleration sensor and a gyro sensor. Furthermore, in the third embodiment, an example in which the inertial sensor 50 is disposed on the wrist of the patient's second hand is described, but this is not limiting. The inertial sensor 50 may be ring-shaped and disposed on a finger of the patient's second hand. In this case, the inertial sensor 50 can also acquire time-series data regarding the grip of the second hand.

[0047] 12 is a block diagram showing an example configuration of a tablet and an inertial sensor. The inertial sensor 50 includes an acceleration sensor 51 and a gyro sensor 52. The acceleration sensor 51 acquires three-axis acceleration data. The gyro sensor 52 acquires three-axis angular velocity data. The inertial sensor 50 outputs the three-axis acceleration data acquired by the acceleration sensor 51 and the three-axis angular velocity data acquired by the gyro sensor 52 to the tablet 20.

[0048] The tablet 20 includes an inertial sensor I / F 28. The inertial sensor I / F 28 is an interface that connects the tablet 20 with the inertial sensor 50. The control unit 21 acquires the three-axis acceleration data and the three-axis angular velocity data output from the inertial sensor 50 via the inertial sensor I / F 28.

[0049] FIG. 13 is an explanatory diagram of a learning model according to the second embodiment. In the second embodiment, an example will be described in which the learning model 342 receives triaxial acceleration data and triaxial angular velocity data as input, and outputs the probability of self-removal based on the input triaxial acceleration data and triaxial angular velocity data. The probability of self-removal is output as a value between 0 and 1. The learning model 342 is configured using, for example, a Transformer. Note that the learning model 342 may be configured using other algorithms such as CNN, RNN, GPT, or LSTN, or may be configured by combining multiple algorithms.

[0050] In the second embodiment, an example will be described in which triaxial acceleration data and triaxial angular velocity data are used in the learning model 342, but this is not limiting. The learning model 342 may output the probability of self-extraction using at least one of the triaxial acceleration data and the triaxial angular velocity data. Furthermore, the learning model 342 may output the probability of self-extraction using data obtained by frequency-transforming the triaxial acceleration data or the triaxial angular velocity data.

[0051] Learning model 342 has an input layer to which triaxial acceleration data and triaxial angular velocity data are input, a middle layer that extracts features from the input time-series data, and an output layer that outputs the probability of self-extraction based on the calculation results of the middle layer. The input layer has input nodes to which triaxial acceleration data and triaxial angular velocity data are input. The middle layer calculates output values ​​based on the time-series data input via the input layer using various functions, thresholds, etc. The output layer has output nodes that output the probability of self-extraction. With the above-mentioned configuration, learning model 342 outputs the probability of self-extraction when triaxial acceleration data and triaxial angular velocity data are input.

[0052] In FIG. 13, when the control unit 31 inputs the three-axis acceleration data and the three-axis angular velocity data to the learning model 342, the learning model 342 outputs a probability of "0.8" of self-removal. If the probability of self-removal is equal to or greater than a predetermined threshold (e.g., 0.75), the control unit 31 acquires "self-removal will occur" as the estimated result. If the probability of self-removal is less than the predetermined threshold (e.g., 0.75), the control unit 31 acquires "self-removal will not occur" as the estimated result. In FIG. 13, the probability of self-removal of "0.8" is equal to or greater than the predetermined threshold, so the control unit 31 acquires "self-removal will occur" as the estimated result.

[0053] FIG. 14 is an explanatory diagram showing an example of a record layout of a training DB according to the second embodiment. The training DB 343 stores training data used to generate the learning model 342. The training DB 343 includes a time-series data sequence and a self-removal sequence. The time-series data sequence stores time-series data showing signs of self-removal and time-series data showing no signs of self-removal. The time-series data showing signs of self-removal includes, for example, time-series data showing a state in which the patient frequently touches the fixation part at the puncture site and time-series data showing a state in which the patient's body is moving vigorously. The time-series data showing signs of self-removal includes triaxial acceleration data and triaxial angular velocity data showing signs of self-removal. The time-series data showing no signs of self-removal includes, for example, time-series data showing a state in which the patient does not touch the fixation part at the puncture site and time-series data showing a state in which the patient's body is moving calmly. The time-series data showing no signs of self-removal includes triaxial acceleration data and triaxial angular velocity data showing no signs of self-removal. The self-removal column stores information indicating whether or not there is a sign of self-removal ("1": self-removal will occur / "0": self-removal will not occur). In the example of FIG. 14, time-series data showing signs of leading to self-removal is associated with information indicating whether or not there is a sign of self-removal ("1": self-removal will occur) and stored. Also, in the example of FIG. 14, time-series data not showing signs of leading to self-removal is associated with information indicating whether or not the patient has self-removed ("0": self-removal will not occur).

[0054] A method for generating the learning model 342 according to the second embodiment will be described. The control unit 31 reads out from the training DB 343 a plurality of training data sets in which time-series data and information indicating the presence or absence of signs of self-removal are associated. When time-series data for training is input to the learning model 342, the control unit 31 trains the learning model 342 so that it outputs the correct value (0 or 1) of the training data. The control unit 31 optimizes parameters used in the calculation process in the intermediate layer so that the output value from the output layer of the learning model 342 approaches the correct value of the training data. The parameters are, for example, the weights (coupling coefficients) of the nodes or the coefficients of the activation functions used in each node. The parameter optimization method is, for example, the backpropagation method or the steepest descent method. The control unit 31 stores the learning model 342 generated by the above process in the mass storage unit 34.

[0055] The processing of the second embodiment will be described. The acceleration sensor 51 acquires triaxial acceleration data. The gyro sensor 52 acquires triaxial angular velocity data. The inertial sensor 50 outputs the triaxial acceleration data acquired by the acceleration sensor 51 and the triaxial angular velocity data acquired by the gyro sensor 52 to the tablet 20. The control unit 21 acquires the triaxial acceleration data and the triaxial angular velocity data output from the inertial sensor 50 via the inertial sensor I / F 28. The control unit 21 divides the acquired triaxial acceleration data into triaxial acceleration data for a predetermined time period (e.g., 30 seconds). The control unit 21 divides the triaxial acceleration data so that the previously divided triaxial acceleration data and the next triaxial acceleration data to be divided include overlapping portions. Specifically, when dividing the acquired triaxial acceleration data into 30-second triaxial acceleration data, the control unit 21 divides the triaxial acceleration data so that the previously divided triaxial acceleration data and the next triaxial acceleration data to be divided include triaxial acceleration data for 15 seconds, where the previously divided triaxial acceleration data and the next triaxial acceleration data to be divided overlap. When dividing the acquired triaxial acceleration data, the control unit 21 may divide the triaxial acceleration data so that the previously divided triaxial acceleration data and the next triaxial acceleration data to be divided do not include overlapping portions. The control unit 21 divides the acquired triaxial angular velocity data in the same manner as for the triaxial acceleration data. The control unit 21 sequentially transmits the divided triaxial acceleration data and triaxial angular velocity data, as well as the patient ID, to the server 30.

[0056] The control unit 31 receives the triaxial acceleration data, triaxial angular velocity data, and patient ID transmitted from the tablet 20. The control unit 31 reads the patient's name and room number associated with the patient ID from the patient DB 341. The control unit 31 inputs the received triaxial acceleration data and triaxial angular velocity data into the learning model 342. The control unit 31 acquires the probability of self-removal output from the learning model 342. The control unit 31 determines whether the acquired probability is equal to or greater than a predetermined threshold. If the control unit 31 determines that the acquired probability is not equal to or greater than the predetermined threshold, it again receives the triaxial acceleration data, triaxial angular velocity data, and patient ID transmitted from the tablet 20. If the control unit 31 determines that the acquired probability is equal to or greater than the predetermined threshold, it generates an alert to prevent the patient from self-removing the device using the patient's name and room number. The subsequent processing is similar to that in the first embodiment, and therefore a description thereof will be omitted.

[0057] 15 is a flowchart showing a procedure for generating a learning model according to the second embodiment. The control unit 31 reads out from the training DB 343 a plurality of training data in which video data is associated with information indicating the presence or absence of signs of self-removal (step S501). The control unit 31 uses the read out training data to generate a learning model 342 that receives triaxial acceleration data and triaxial angular velocity data as input and outputs the probability of self-removal (step S502). The control unit 31 stores the generated learning model 342 in the mass storage unit 34 (step S503).

[0058] FIG. 16 is a flowchart showing an example of a processing procedure of the self-removal prevention system according to the second embodiment. In the flowchart shown in FIG. 16, steps S201 to S205 in the processing shown in FIG. 10 are replaced with steps S601 to S605. Descriptions of steps similar to those in FIG. 10 will be omitted. The control unit 31 determines whether or not the triaxial acceleration data, triaxial angular velocity data, and patient ID transmitted from the tablet 20 have been received (step S601). If the control unit 31 determines that the triaxial acceleration data, triaxial angular velocity data, and patient ID have not been received (step S601: NO), the control unit 31 waits until the triaxial acceleration data, triaxial angular velocity data, and patient ID are received. If the control unit 31 determines that the triaxial acceleration data, triaxial angular velocity data, and patient ID have been received (step S601: YES), the control unit 31 reads out the name and room number of the patient associated with the patient ID from the patient DB 341 (step S602). The control unit 31 inputs the received three-axis acceleration data and three-axis angular velocity data to the learning model 342 (step S603). The control unit 31 acquires the probability of self-removal output by the learning model 342 (step S604). The control unit 31 determines whether the acquired probability is equal to or greater than a predetermined threshold (step S605). If the control unit 31 determines that the acquired probability is not equal to or greater than the predetermined threshold (step S605: NO), the control unit 31 returns the process to step S601. If the control unit 31 determines that the acquired probability is equal to or greater than the predetermined threshold (step S605: YES), the control unit 31 proceeds to step S206.

[0059] According to embodiment 2, the self-removal prevention system acquires three-axis acceleration data and three-axis angular velocity data from an inertial sensor attached to the second hand, and inputs the three-axis acceleration data and three-axis angular velocity data into a learning model, thereby outputting the probability of self-removal.

[0060] (Embodiment 3) Embodiment 3 describes an example in which a self-removal prevention system outputs the probability of self-removal by inputting into a learning model first time series data regarding the movement of the first hand obtained by a first inertial sensor attached to the first hand and second time series data regarding the movement of the second hand obtained by a second inertial sensor attached to the second hand.

[0061] FIG. 17 is an explanatory diagram showing an overview of a self-removal prevention system according to a third embodiment. The self-removal prevention system 10 includes a first inertial sensor 60 and a second inertial sensor 70. The first inertial sensor 60 is attached to the wrist of the patient's first hand (left hand in FIG. 17) that has been punctured, using a strap or the like. The second inertial sensor 70 is attached to the wrist of the patient's second hand (right hand in FIG. 17) that is different from the first hand, using a strap or the like. A wearable device may be used instead of the first inertial sensor 60 or the second inertial sensor 70. The first inertial sensor 60, the second inertial sensor 70, and the tablet 20 are connected to each other so as to be able to communicate with each other via short-range wireless communication such as Bluetooth. The tablet 20, the first inertial sensor 60, the second inertial sensor 70, and the patient are associated with each other using a patient ID.

[0062] The first inertial sensor 60 includes an acceleration sensor, a gyro sensor, etc., and acquires first time-series data related to the movement of the patient's first hand. The first time-series data is, for example, time-series data of triaxial acceleration related to the movement of the patient's first hand (hereinafter referred to as first acceleration data) and time-series data of triaxial angular velocity related to the movement of the patient's first hand (hereinafter referred to as first angular velocity data). The first inertial sensor 60 also includes a proximity sensor such as a magnetic proximity sensor, and detects the distance between the first hand and the second hand. The second inertial sensor 70 includes an acceleration sensor, a gyro sensor, etc., and acquires second time-series data related to the movement of the patient's second hand. The second time-series data is, for example, time-series data of triaxial acceleration related to the movement of the patient's second hand (hereinafter referred to as second acceleration data) and time-series data of triaxial angular velocity related to the movement of the patient's second hand (hereinafter referred to as second angular velocity data).

[0063] In the third embodiment, an example will be described in which the first inertial sensor 60 includes an acceleration sensor and a gyro sensor, and the second inertial sensor 70 includes an acceleration sensor and a gyro sensor, but this is not limiting. The first inertial sensor 60 and the second inertial sensor 70 may each include at least one of an acceleration sensor and a gyro sensor.

[0064] 18 is a block diagram showing an example configuration of the tablet, the first inertial sensor, and the second inertial sensor. The first inertial sensor 60 includes a first acceleration sensor 61, a first gyro sensor 62, and a proximity sensor 63. The first acceleration sensor 61 acquires first acceleration data. The first gyro sensor 62 acquires first angular velocity data. The first inertial sensor 60 outputs the first acceleration data acquired by the first acceleration sensor 61 and the first angular velocity data acquired by the first gyro sensor 62 to the tablet 20.

[0065] The proximity sensor 63 is, for example, a magnetic proximity sensor, and detects the distance between the first hand and the second hand using a magnet (not shown) built into the second inertial sensor 70. When the distance between the first hand and the second hand is within a predetermined range (for example, within 10 cm), the proximity sensor 63 outputs a notification (hereinafter referred to as a proximity notification) to the tablet 20 that the first hand and the second hand are in proximity. The proximity sensor 63 may be, for example, an inductive proximity sensor or a capacitive proximity sensor. The second inertial sensor 70 includes a second acceleration sensor 71 and a second gyro sensor 72. The second acceleration sensor 71 acquires second acceleration data. The second gyro sensor 72 acquires second angular velocity data. The second inertial sensor 70 outputs the second acceleration data acquired by the second acceleration sensor 71 and the second angular velocity data acquired by the second gyro sensor 72 to the tablet 20.

[0066] In the third embodiment, the first inertial sensor 60 includes the proximity sensor 63, but this is not limiting. The second inertial sensor 70 may include the proximity sensor 63. In this case, a magnet used in conjunction with the proximity sensor 63 is built into the first inertial sensor 60. The proximity sensor 63 may also be placed on the patient's first hand independently of the first inertial sensor 60. In this case, a magnet used in conjunction with the proximity sensor 63 is placed on the patient's second hand independently of the second inertial sensor 70. The proximity sensor 63 may also be placed on the patient's second hand independently of the second inertial sensor 70. In this case, a magnet used in conjunction with the proximity sensor 63 is placed on the patient's first hand independently of the first inertial sensor 60. When the proximity sensor 63 and the magnet used in conjunction with the proximity sensor 63 are independent of the first inertial sensor 60 or the second inertial sensor 70, they are placed on the first or second hand using a strap or the like.

[0067] The control unit 21 acquires the first acceleration data and the first angular velocity data output from the first inertial sensor 60 via the inertial sensor I / F 28. The control unit 21 also acquires the second acceleration data and the second angular velocity data output from the second inertial sensor 70 via the inertial sensor I / F 28.

[0068] FIG. 19 is an explanatory diagram of a learning model according to the third embodiment. In the third embodiment, an example will be described in which a learning model 342 receives first time series data and second time series data as input and outputs a probability of self-removal based on the input first time series data and second time series data. The probability of self-removal is output as a value between 0 and 1. The first time series data includes first acceleration data and first angular velocity data. The second time series data includes second acceleration data and second angular velocity data. The probability of self-removal is output as a value between 0 and 1. The learning model 342 is configured using, for example, a Transformer. Note that the learning model 342 may be configured using other algorithms such as CNN, RNN, GPT, or LSTN, or may be configured by combining multiple algorithms. In the third embodiment, an example will be described in which the first time series data includes first acceleration data and first angular velocity data, and the second time series data includes second acceleration data and second angular velocity data, but this is not limiting. The first time series data may include at least one of the first acceleration data and the first angular velocity data. The second time-series data may include at least one of the second acceleration data and the second angular velocity data.

[0069] The learning model 342 has an input layer to which the first time series data and the second time series data are input, a middle layer that extracts features from the input first time series data and the second time series data, and an output layer that outputs the self-extraction probability based on the calculation results of the middle layer. The input layer has input nodes to which the first time series data and the second time series data are input. The middle layer calculates output values ​​based on the first time series data and the second time series data input via the input layer using various functions, thresholds, etc. The output layer has an output node that outputs the self-extraction probability. With the above-mentioned configuration, the learning model 342 outputs the self-extraction probability when the first time series data and the second time series data are input.

[0070] In FIG. 19, when the control unit 31 inputs the first time series data and the second time series data received from the tablet 20 into the learning model 342, the learning model 342 outputs a probability of "0.8" of self-removal. If the probability of self-removal is equal to or greater than a predetermined threshold (e.g., 0.75), the control unit 31 acquires "self-removal will occur" as the estimated result. If the probability of self-removal is less than the predetermined threshold (e.g., 0.75), the control unit 31 acquires "self-removal will not occur" as the estimated result. In FIG. 19, the probability of self-removal of "0.8" is equal to or greater than the predetermined threshold, so the control unit 31 acquires "self-removal will occur" as the estimated result.

[0071] FIG. 20 is an explanatory diagram illustrating an example of a record layout of a training DB according to the third embodiment. The training DB 343 stores training data used to generate the learning model 342. The training DB 343 includes a time-series data sequence and a self-removal sequence. The time-series data sequence stores first time-series data indicating signs of self-removal and second time-series data indicating signs of self-removal. The time-series data sequence also stores first time-series data not indicating signs of self-removal and second time-series data not indicating signs of self-removal. The self-removal sequence stores information indicating the presence or absence of signs of self-removal (“1”: self-removal will occur / “0”: self-removal will not occur). In the example of FIG. 20, the first time-series data indicating signs of self-removal and the second time-series data indicating signs of self-removal are stored in association with information indicating the presence or absence of signs of self-removal (“1”: self-removal will occur). In the example of Figure 20, the first time series data showing no signs of self-removal and the second time series data showing no signs of self-removal are stored in association with information indicating whether or not the patient will self-removal ("0": no self-removal).

[0072] A method for generating the learning model 342 according to the third embodiment will be described. The control unit 31 reads out from the training DB 343 a plurality of training data sets in which first and second time series data are associated with information indicating the presence or absence of signs of self-removal. When the first and second time series data for training are input to the learning model 342, the control unit 31 trains the learning model 342 so that it outputs a correct value (0 or 1) of the training data. The control unit 31 optimizes parameters used in the calculation process in the intermediate layer so that the output value from the output layer of the learning model 342 approaches the correct value of the training data. The parameters are, for example, the weights (coupling coefficients) of the nodes or the coefficients of the activation functions used in each node. The parameter optimization method is, for example, the backpropagation algorithm or the steepest descent algorithm. The control unit 31 stores the learning model 342 generated by the above process in the mass storage unit 34.

[0073] The processing of the third embodiment will be described. The first acceleration sensor 61 acquires first acceleration data. The first gyro sensor 62 acquires first angular velocity data. The first inertial sensor 60 outputs first time series data including the first acceleration data and the first angular velocity data to the tablet 20. The second acceleration sensor 71 acquires second acceleration data. The second gyro sensor 72 acquires second angular velocity data. The second inertial sensor 70 outputs second time series data including the second acceleration data and the second angular velocity data to the tablet 20. The control unit 21 acquires the first time series data output from the first inertial sensor 60 via the inertial sensor I / F 28. The control unit 21 also acquires the second time series data output from the second inertial sensor 70 via the inertial sensor I / F 28.

[0074] The control unit 21 divides the acquired first time series data into first time series data of a predetermined time period (e.g., 30 seconds). The control unit 21 divides the first time series data so that the previously divided first time series data and the next divided first time series data include an overlapping portion. Specifically, when dividing the acquired first time series data into 30-second first time series data, the control unit 21 divides the first time series data so that the previously divided first time series data and the next divided first time series data include a 15-second first time series data overlapping portion. The control unit 21 may also divide the first time series data so that the previously divided first time series data and the next divided first time series data do not include an overlapping portion. The control unit 21 divides the acquired second time series data in the same manner as the first time series data. The control unit 21 sequentially transmits the divided first time series data, the divided second time series data, and the patient ID to the server 30.

[0075] The control unit 31 receives the first time series data, the second time series data, and the patient ID transmitted from the tablet 20. The control unit 31 reads the patient's name and room number associated with the patient ID from the patient DB 341. The control unit 31 inputs the received first time series data and second time series data into the learning model 342. The control unit 31 acquires the probability of self-removal output by the learning model 342. The control unit 31 determines whether the acquired probability is equal to or greater than a predetermined threshold. If the control unit 31 determines that the acquired probability is not equal to or greater than the predetermined threshold, it again receives the first time series data, the second time series data, and the patient ID transmitted from the tablet 20. If the control unit 31 determines that the acquired probability is equal to or greater than the predetermined threshold, it determines whether a proximity notification has been received from the tablet 20. If the control unit 31 determines that a proximity notification has not been received from the tablet 20, it again receives the first time series data, the second time series data, and the patient ID transmitted from the tablet 20. When the control unit 31 determines that it has received a proximity notification from the tablet 20, it generates an alert for the patient and an alert for the medical staff using the patient's name and room number. The subsequent processing is the same as in the first embodiment, and therefore a description thereof will be omitted.

[0076] 21 is a flowchart showing a procedure for generating a learning model according to the third embodiment. The control unit 31 reads out from the training DB 343 a plurality of training data in which first and second time series data are associated with information indicating the presence or absence of signs of self-removal (step S701). The control unit 11 uses the read out training data to generate a learning model 342 that receives the first and second time series data as input and outputs the probability of self-removal (step S702). The control unit 11 stores the generated learning model 342 in the mass storage unit 34 (step S703).

[0077] FIG. 22 is a flowchart showing an example of a processing procedure of the self-removal prevention system according to the third embodiment. In the flowchart shown in FIG. 22, steps S201 to S205 in the processing shown in FIG. 10 are replaced with steps S801 to S806. Descriptions of steps similar to those in FIG. 10 will be omitted. The control unit 31 determines whether the first time series data, the second time series data, and the patient ID transmitted from the tablet 20 have been received (step S801). If the control unit 31 determines that the first time series data, the second time series data, and the patient ID have not been received (step S801: NO), the control unit 31 waits until the first time series data, the second time series data, and the patient ID are received. If the control unit 31 determines that the first time series data, the second time series data, and the patient ID have been received (step S801: YES), the control unit 31 reads out the name and room number of the patient associated with the patient ID from the patient DB 341 (step S802).

[0078] The control unit 31 inputs the received first time-series data and second time-series data to the learning model 342 (step S803). The control unit 31 acquires the probability of self-removal output by the learning model 342 (step S804). The control unit 31 determines whether the acquired probability is equal to or greater than a predetermined threshold (step S805). If the control unit 31 determines that the acquired probability is not equal to or greater than the predetermined threshold (step S805: NO), the control unit 31 returns the process to step S801. If the control unit 31 determines that the acquired probability is equal to or greater than the predetermined threshold (step S805: YES), the control unit 31 determines whether a proximity notification has been received from the tablet 20 (step S806). If the control unit 31 determines that a proximity notification has not been received from the tablet 20 (step S806: NO), the control unit 31 returns the process to step S801. If the control unit 31 determines that a proximity notification has been received from the tablet 20 (step S806: YES), the control unit 31 proceeds to step S206.

[0079] (Variation) In the third embodiment, an example has been described in which the proximity sensor 63 is used to detect whether the first hand and the second hand are in proximity to each other, but this is not limiting. The self-removal prevention system may use a distance sensor instead of the proximity sensor 63. In this case, the distance sensor is communicatively connected to the tablet 20. The distance sensor emits laser light toward an object (e.g., the wrist of the first hand and the wrist of the second hand) and measures the distance to the object based on the time it takes for the reflected laser light to be received. The distance sensor may use ultrasound instead of laser light. The distance sensor uses laser light to measure the distance from the tablet 20 to the first hand (hereinafter referred to as the first distance) and the distance from the tablet 20 to the second hand (hereinafter referred to as the second distance). In addition, the distance sensor measures the angle between the first distance and the second distance in conjunction with measuring the first distance and the second distance. The distance sensor transmits the first distance, the second distance, and the angle between the first distance and the second distance measured by the distance sensor to the tablet 20.

[0080] The tablet 20 acquires the first distance, the second distance, and the angle formed by the first distance and the second distance transmitted from the distance sensor. The control unit 21 calculates the average value of the first distance, the average value of the second distance, and the average value of the angle formed by the first distance and the second distance over a predetermined period (e.g., 5 seconds). The control unit 21 calculates the distance from the first hand to the second hand based on the calculated average values ​​of the first distance, the average value of the second distance, and the average value of the angle formed by the first distance and the second distance, for example, by using the cosine law. The distance from the first hand to the second hand is, for example, the distance from the wrist of the first hand to the wrist of the second hand, or the distance from the thumb of the first hand to the thumb of the second hand. The control unit 21 transmits the distance from the first hand to the second hand to the server 30 in association with the first time-series data, the second time-series data, and the patient ID.

[0081] The control unit 31 receives the first time series data, the second time series data, the patient ID, and the distance from the first hand to the second hand transmitted from the tablet 20. The control unit 31 reads the patient's name and room number associated with the patient ID from the patient DB 341. The control unit 31 inputs the received first time series data and second time series data into the learning model 342. The control unit 31 acquires the probability of self-removal output by the learning model 342. The control unit 31 determines whether the acquired probability is equal to or greater than a predetermined threshold. If the control unit 31 determines that the acquired probability is not equal to or greater than the predetermined threshold, it receives the first time series data and the second time series data transmitted from the tablet 20 again. If the control unit 31 determines that the acquired probability is equal to or greater than the predetermined threshold, it determines whether the distance from the first hand to the second hand is within a predetermined range (for example, within 10 cm). If the control unit 31 determines that the distance from the first hand to the second hand is not within the predetermined range, it again receives the first time series data and the second time series data transmitted from the tablet 20. If the control unit 31 determines that the distance from the first hand to the second hand is within the predetermined range, it generates an alert using the patient's name and room number to prevent the patient from removing the device themselves. The subsequent processing is the same as in embodiment 3, so a description thereof will be omitted.

[0082] According to the third embodiment, the self-removal prevention system can acquire first time-series data regarding the movement of the first hand obtained by the first sensor attached to the first hand.

[0083] According to the third embodiment, the self-removal prevention system can acquire second time-series data regarding the movement of the second hand obtained by an inertial sensor attached to the second hand.

[0084] According to the third embodiment, the self-removal prevention system can output the probability of self-removal by inputting the acquired first time-series data and second time-series data into the learning model.

[0085] According to the third embodiment, the self-removal prevention system can output an alert when the distance from the first hand to the second hand is within a predetermined range and the probability of self-removal is equal to or greater than a predetermined threshold.

[0086] (Embodiment 4) In the fourth embodiment, an example will be described in which the self-removal prevention system changes the predetermined threshold or the alert depending on the patient's condition.

[0087] FIG. 23 is an explanatory diagram showing an example of a record layout of a patient DB according to the fourth embodiment. The patient DB 341 includes a medical condition column. The medical condition column stores the medical condition of the patient. The medical condition of the patient is, for example, "none (healthy)", "dementia", or "delirium / disturbed consciousness". In the record of patient ID K001 in FIG. 23, the name "A", room number "Room 301", and medical condition "none (healthy)" are stored.

[0088] First, an example will be described in which the self-removal prevention system changes the predetermined threshold value according to the patient's condition. The control unit 31 receives video data and a patient ID transmitted from the tablet 20. The control unit 31 reads the patient's name, room number, and condition associated with the patient ID from the patient DB 341. The control unit 31 determines whether the read patient's condition is "healthy." If the control unit 31 determines that the read patient's condition is "healthy," the control unit 31 inputs the received video data into the learning model 342 without changing the predetermined threshold value. The subsequent processing when the control unit 31 determines that the patient's condition is "healthy" is the same as in the first embodiment. If the control unit 31 determines that the read patient's condition is not "healthy," the control unit 31 changes the predetermined threshold value. For example, the control unit 31 downwardly adjusts the predetermined threshold value from "0.75" to "0.6." The control unit 31 inputs the received video data into the learning model 342. The control unit 31 acquires the probability of self-removal output by the learning model 342. The control unit 31 determines whether the acquired probability is equal to or greater than the changed predetermined threshold. If the control unit 31 determines that the acquired probability is not equal to or greater than the changed predetermined threshold, it receives the video data and patient ID transmitted from the tablet 20 again. If the control unit 31 determines that the acquired probability is equal to or greater than the changed predetermined threshold, it generates an alert for the patient and an alert for the medical staff using the patient's name and room number. The subsequent processing is the same as in embodiment 1, and therefore description thereof will be omitted.

[0089] Next, an example will be described in which the self-removal prevention system changes alerts depending on the patient's condition. The control unit 31 receives video data and a patient ID transmitted from the tablet 20. The control unit 31 reads the patient's name, room number, and condition associated with the patient ID from the patient DB 341. The control unit 31 inputs the received video data into the learning model 342. The control unit 31 acquires the probability of self-removal output by the learning model 342. The control unit 31 determines whether the acquired probability is equal to or greater than a predetermined threshold. If the control unit 31 determines that the acquired probability is not equal to or greater than the predetermined threshold, it again receives the video data and the patient ID transmitted from the tablet 20. If the control unit 31 determines that the acquired probability is equal to or greater than the predetermined threshold, it generates an alert for the patient and an alert for medical personnel using the patient's name and room number. The control unit 31 determines whether the patient's condition is "healthy." If the control unit 31 determines that the read patient's condition is "healthy," it transmits an alert for the patient to the tablet 20. Furthermore, if the control unit 31 determines that the read-out medical condition of the patient is "healthy", it transmits an alert for medical staff to the smartphone 40. The subsequent processing when the patient's medical condition is determined to be "healthy" is the same as in embodiment 1, and therefore description thereof will be omitted. If the control unit 31 determines that the read-out medical condition of the patient is not "healthy", it modifies the generated alert for the patient and alert for medical staff. The control unit 31 transmits the modified alert for the patient to the tablet 20. Furthermore, the control unit 31 transmits the modified alert for medical staff to the smartphone 40.

[0090] The control unit 21 receives a patient alert transmitted from the server 30. Upon receiving the patient alert from the server 30, the control unit 21 immediately displays the received patient alert on the display unit 24. FIG. 24A is an explanatory diagram showing an example screen of a changed patient alert. The screen d3 in FIG. 24A includes a patient information field d31, a patient message field d32, and a confirmation button b33. The content of the patient information field d31 is the same as that in the first embodiment. The patient message field d32 in FIG. 24A displays "Do not remove the IV drip." The confirmation button b33 is a button for instructing the end of the display of the screen d3. In parallel with the display of the screen d3, the control unit 21 repeatedly outputs a warning sound (e.g., a beep) and the content of the patient message field d32 as audio from the speaker 25. The control unit 21 determines whether or not selection of the confirmation button b33 has been accepted. If the control unit 21 determines that selection of the confirmation button b33 has been accepted, the control unit 21 ends the processing. If the control unit 21 determines that the selection of the confirmation button b33 has not been accepted, it continues to display the screen d3 and output the sound.

[0091] The control unit 41 receives a medical worker alert sent from the server 30. Upon receiving the medical worker alert from the server 30, the control unit 41 immediately displays the received medical worker alert on the display unit 44. FIG. 24B is an explanatory diagram showing an example of a screen displaying a changed medical worker alert. The screen d4 in FIG. 24B includes a medical worker message field d41 and a confirmation button b43. The confirmation button b43 is a button for instructing the display of the screen d4 to end. The medical worker message field d41 in FIG. 24B displays the following: "Mr. A in Room 301 is attempting to remove his IV drip himself. Please rush to the scene immediately. Mr. A has dementia, so please communicate more slowly and in shorter words than usual." The control unit 41 determines whether selection of the confirmation button b43 has been accepted. If the control unit 41 determines that selection of the confirmation button b43 has been accepted, the control unit 41 terminates the process. If the control unit 41 determines that the selection of the confirmation button b43 has not been accepted, it continues to display the screen d4.

[0092] In the fourth embodiment, an example has been described in which the self-removal prevention system changes the predetermined threshold or the alert depending on the patient's condition, but this is not limiting. The self-removal prevention system may change the predetermined threshold and the alert depending on the patient's condition.

[0093] FIG. 25 is a flowchart showing an example of a processing procedure of the self-removal prevention system according to the fourth embodiment. The flowchart shown in FIG. 25 is a flowchart for changing the predetermined threshold value according to the condition of a patient. In the flowchart shown in FIG. 25, steps S201 to S206 in the processing shown in FIG. 10 are replaced with steps S901 to S908. Descriptions of steps similar to those in FIG. 10 will be omitted. The control unit 31 determines whether video data and a patient ID have been received from the tablet 20 (step S901). If the control unit 31 determines that the video data and the patient ID have not been received from the tablet 20 (step S901: NO), the control unit 31 waits until the video data and the patient ID are received. If the control unit 31 determines that the video data and the patient ID transmitted from the tablet 20 have been received (step S901: YES), the control unit 31 reads the name, room number, and condition of the patient associated with the patient ID from the patient DB 341 (step S902).

[0094] The control unit 31 determines whether the read-out medical condition of the patient is "healthy" (step S903). If the control unit 31 determines that the medical condition of the patient is "healthy" (step S903: YES), the control unit 31 proceeds to step S203. If the control unit 31 determines that the medical condition of the patient is not "healthy" (step S903: NO), the control unit 31 changes the predetermined threshold (step S904). The control unit 31 inputs the received video data into the learning model 342 (step S905). The control unit 31 acquires the probability of self-removal output by the learning model 342 (step S906). The control unit 31 determines whether the acquired probability is equal to or greater than the changed predetermined threshold (step S907). If the control unit 31 determines that the acquired probability is not equal to or greater than the changed predetermined threshold (step S907: NO), the control unit 31 returns to step S901. If the control unit 31 determines that the acquired probability is equal to or greater than the changed predetermined threshold (step S907: YES), the control unit 31 generates an alert for the patient and an alert for the medical staff using the patient's name and room number (step S908). The control unit 31 proceeds to step S207.

[0095] FIG. 26 is a flowchart showing an example of a processing procedure of the self-removal prevention system according to the fourth embodiment. The flowchart shown in FIG. 26 is a flowchart for changing an alert according to the patient's condition. The control unit 31 determines whether or not the video data and the patient ID transmitted from the tablet 20 have been received (step S1001). If the control unit 31 determines that the video data and the patient ID have not been received from the tablet 20 (step S1001: NO), the control unit 31 waits until the video data and the patient ID are received. If the control unit 31 determines that the video data and the patient ID transmitted from the tablet 20 have been received (step S1001: YES), the control unit 31 reads the name, room number, and condition of the patient associated with the patient ID from the patient DB 341 (step S1002). The control unit 31 inputs the received video data into the learning model 342 (step S1003). The control unit 31 acquires the probability of self-removal output by the learning model 342 (step S1004). The control unit 31 determines whether the acquired probability is equal to or greater than a predetermined threshold (step S1005). If the control unit 31 determines that the acquired probability is not equal to or greater than the predetermined threshold (step S1005: NO), the control unit 31 returns the process to step S1001. If the control unit 31 determines that the acquired probability is equal to or greater than the predetermined threshold (step S1005: YES), the control unit 31 generates an alert for the patient and an alert for the medical staff using the patient's name and room number (step S1006).

[0096] The control unit 31 determines whether the patient's condition is "healthy" (step S1007). If the control unit 31 determines that the read-out patient's condition is "healthy" (step S1007: YES), the process proceeds to step S207. If the control unit 31 determines that the read-out patient's condition is not "healthy" (step S1007: NO), the control unit 31 changes the alert for the patient and the alert for the medical staff generated in step S1006 (step S1008). The control unit 31 transmits the changed alert for the patient to the tablet 20 (step S1009). The control unit 31 transmits the changed alert for the medical staff to the smartphone 40 (step S1010).

[0097] The control unit 21 receives the patient alert transmitted from the server 30 (step S1101). Upon receiving the patient alert from the server 30, the control unit 21 immediately displays the patient alert on the display unit 24 (step S1102). The control unit 21 determines whether or not selection of the confirmation button has been accepted (step S1103). If the control unit 21 determines that selection of the confirmation button has been accepted (step S1103: YES), it ends the process. If the control unit 21 determines that selection of the confirmation button has not been accepted (step S1103: NO), it returns the process to step S1102.

[0098] The control unit 41 receives an alert for medical personnel transmitted from the server 30 (step S1201). Upon receiving the alert for medical personnel from the server 30, the control unit 41 immediately displays the received alert for medical personnel on the display unit 44 (step S1202). The control unit 41 determines whether or not selection of the confirmation button has been accepted (step S1203). If the control unit 41 determines that selection of the confirmation button has been accepted (step S1203: YES), it ends the processing. If the control unit 41 determines that selection of the confirmation button has not been accepted (step S1203: NO), it returns the processing to step S1202.

[0099] According to the fourth embodiment, the self-removal prevention system can change the predetermined threshold or alert depending on the patient's condition.

[0100] (Embodiment 5) In the fifth embodiment, an example is described in which the self-removal prevention system outputs a command to give a stimulus to the patient to prevent self-removal. The stimulus to prevent self-removal is, for example, vibration or light. In the fifth embodiment, an example of vibration is described as the stimulus to prevent self-removal.

[0101] FIG. 27 is a block diagram showing an example of the configuration of a tablet and an inertial sensor according to the fifth embodiment. The inertial sensor 50 includes a vibration generating unit 53. The vibration generating unit 53 includes a vibrator that generates vibrations to prevent the patient from removing the device themselves. The vibration generating unit 53 generates vibrations to prevent the patient from removing the device themselves, based on an instruction from the tablet 20, using the vibrator. The vibration generating unit 53 may be disposed in a bed sensor (not shown). In this case, the bed sensor includes a communication unit capable of wireless communication with the tablet 20 or the server 30. The bed sensor is, for example, a sheet-like sensor that is disposed between the bed mattress and the sheet.

[0102] The processing of the fifth embodiment will be described. When the control unit 31 determines that the probability of self-removal acquired by the learning model 342 is equal to or greater than a predetermined threshold, the control unit 31 generates an alert for the patient using the patient's name and room number. When the control unit 31 determines that the acquired probability is equal to or greater than a predetermined threshold, the control unit 31 generates a command to give the patient a stimulus to prevent self-removal (hereinafter referred to as a self-removal prevention command). The control unit 31 transmits the patient alert and the self-removal prevention command to the tablet 20.

[0103] The control unit 21 receives the patient alert and the self-removal prevention command transmitted from the server 30. The control unit 21 displays the received patient alert on the display unit 24. The control unit 21 causes the vibration generation unit 53 to generate vibrations in parallel with displaying the patient alert.

[0104] In the fifth embodiment, an example has been described in which vibration is used as a stimulus to prevent self-removal, but this is not limiting. Light may also be used as a stimulus to prevent self-removal. In this case, the tablet 20 is provided with a lamp. When the control unit 21 receives a patient alert and a self-removal prevention command transmitted from the server 30, the control unit 21 irradiates light for preventing self-removal from the lamp in parallel with displaying the patient alert.

[0105] Fig. 28 is a flowchart showing an example of the processing procedure of the self-removal prevention system according to the fifth embodiment. In the flowchart shown in Fig. 28, step S206 and the subsequent steps in the processing shown in Fig. 10 are replaced with steps S1301 to S1303 and steps S1401 to S1403. Explanation of the same steps as in Fig. 10 will be omitted. When the control unit 31 determines that the acquired probability is equal to or greater than a predetermined threshold (step S205: YES), the control unit 31 generates an alert for the patient using the patient's name and room number (step S1301). The control unit 31 generates a self-removal prevention command (step S1302). The control unit 31 transmits the alert for the patient and the self-removal prevention command to the tablet 20 (step S1303).

[0106] The control unit 21 receives the patient alert and self-removal prevention command transmitted from the server 30 (step S1401). The control unit 21 displays the received patient alert on the display unit 24 (step S1402). In parallel with displaying the patient alert, the control unit 11 causes the vibration generating unit 53 to generate vibrations (step S1403).

[0107] According to the fifth embodiment, the self-removal prevention system can output a self-removal prevention command when the probability of self-removal is equal to or greater than a predetermined threshold.

[0108] (Embodiment 6) In the sixth embodiment, an example will be described in which the self-removal prevention system outputs a second alert when it detects self-removal by a patient. Methods for detecting self-removal include, for example, a camera or an RFID tag. In the sixth embodiment, a method using a camera will be described.

[0109] The storage unit 22 of the tablet 20 stores sample data (hereinafter referred to as sample data) in advance, which indicates self-removal by the patient. The sample data includes, for example, an image of blood or the like leaking due to self-removal of the infusion, and an image of the puncture site and the separation of the infusion. When the video data acquired by the camera 27 contains data similar to the sample data, the control unit 21 transmits a notification (hereinafter referred to as self-removal notification) indicating that the patient has self-removed the infusion to the server 30. The control unit 21 may use a known machine learning model or the like to determine whether the video data acquired by the camera 27 contains data similar to the sample data (whether the patient has self-removed the infusion).

[0110] The processing of the sixth embodiment will be described. The control unit 31 receives video data and a patient ID transmitted from the tablet 20. The control unit 31 reads the patient's name and room number associated with the patient ID from the patient DB 341. The control unit 31 inputs the received video data into the learning model 342. The control unit 31 acquires the probability of self-removal output from the learning model 342. The control unit 31 determines whether the acquired probability is equal to or greater than a predetermined threshold. If the control unit 31 determines that the acquired probability is equal to or greater than the predetermined threshold, it generates an alert for the patient and an alert for the medical staff using the read patient's name and room number. The subsequent processing when it is determined that the acquired probability is equal to or greater than the predetermined threshold is the same as in the first embodiment, and therefore description thereof will be omitted. If the control unit 31 determines that the acquired probability is not equal to or greater than the predetermined threshold, it determines whether a self-removal notification has been received from the tablet 20. If the control unit 31 determines that a self-removal notification has not been received from the tablet 20, it again receives the video data and the patient ID transmitted from the tablet 20. When the control unit 31 determines that a self-removal notification has been received from the tablet 20, the control unit 31 generates a second alert. The second alert is, for example, an alert indicating that the patient has performed self-removal. The control unit 31 transmits the second alert to the smartphone 40.

[0111] The control unit 41 receives the second alert sent from the server 30. The control unit 41 displays the received second alert on the display unit 44. For example, a message to medical staff, "Mr. A in room 301 has removed his IV drip himself. Please take immediate action," is displayed on the screen of the second alert (not shown).

[0112] In the sixth embodiment, the patient's self-removal is detected using a camera, but this is not limited to this. The patient's self-removal may also be detected using an RFID tag. In this case, the RFID tag is placed at the puncture site. The RFID tag reader / writer is placed at the bottom of the patient's bed. The RFID tag reader / writer is communicably connected to the tablet 20. The control unit 31 determines that the patient has not self-removed the catheter if the RFID tag and the RFID tag reader / writer can communicate with each other. On the other hand, the control unit 31 determines that the patient has self-removed the catheter if the RFID tag and the RFID tag reader antenna cannot communicate with each other. If the control unit 31 determines that the patient has self-removed the catheter, it sends a self-removal notification to the server 30.

[0113] FIG. 29 is a flowchart showing an example of the processing procedure of the self-removal prevention system according to the sixth embodiment. In the flowchart shown in FIG. 29, step S204 and subsequent steps in the processing shown in FIG. 10 are replaced with steps S1501 to S1504 and steps S1601 to S1602. Explanation of steps similar to those in the first embodiment will be omitted. The control unit 31 determines whether the acquired probability is equal to or greater than a predetermined threshold (step S1501). If the control unit 31 determines that the acquired probability is equal to or greater than the predetermined threshold (step S1501: YES), the control unit 31 proceeds to step S206. If the control unit 31 determines that the acquired probability is not equal to or greater than the predetermined threshold (step S1501: NO), the control unit 31 determines whether a self-removal notification has been received from the tablet 20 (step S1502). If the control unit 31 determines that a self-removal notification has not been received from the tablet 20 (step S1502: NO), the control unit 31 returns to step S201. If the control unit 31 determines that a self-removal notification has been received from the tablet 20 (step S1502: YES), the control unit 31 generates a second alert (step S1503) and transmits the second alert to the smartphone 40 (step S1504).

[0114] The control unit 41 receives the second alert transmitted from the server 30 (step S1601). The control unit 41 displays the received second alert on the display unit 44 (step S1602).

[0115] According to the sixth embodiment, the self-removal prevention system can output a second alert when it detects self-removal by the patient.

[0116] The features described in each of the above embodiments can be combined with each other. Furthermore, the independent claims and dependent claims described in the claims can be combined with each other in any and all combinations, regardless of the reference format. Furthermore, the claims use a format in which a claim references two or more other claims (multiple claim format), but this is not limited to this. Multiple claims (multi-multi claims) that reference at least one other multiple claim may also be used.

[0117] The embodiments disclosed herein are to be considered in all respects as illustrative and not restrictive. The scope of the present invention is defined by the claims, not by the above meaning, and is intended to include all modifications within the meaning and scope of the claims. [Explanation of symbols]

[0118] 10 Self-removal prevention system 20 Information processing device (tablet) 21 Control section 22 Memory section 22P control program 23 Communications Department 24 Display section 25 speakers 26 Input section 27 Camera 30 Information processing device (server) 31 Control Unit 32 Storage section 32P control program 33 Communications Department 34 Mass storage 341 Patient DB 342 Learning Model 343 Training DB 35 Reading unit 1a Portable storage media 40 Information processing device (smartphone) 41 Control Unit 42 Storage section 42P control program 43 Communications Department 44 Display section 50 Inertial Sensor 51 Acceleration sensor 52 Gyro sensor 53 Vibration generating unit 60 First inertial sensor 61 First acceleration sensor 62 First gyro sensor 63 Proximity Sensor 70 Second inertial sensor 71 Second acceleration sensor 72 Second gyro sensor

Claims

1. time-series data relating to the movement of a second hand different from the first hand being punctured is acquired; The learning model outputs symptom information regarding signs of self-withdrawal of IV drips when time series data is input. By inputting the acquired time series data, symptom information is output. A program that causes a computer to perform a process.

2. acquiring video data of the puncture site and the movement of the second hand captured by a camera; The video data is input to the learning model, and the symptom information is output. The program according to claim 1.

3. Acquire the video data that does not include a facial image of the patient; The acquired video data is input to the learning model, and the symptom information is output. The program according to claim 2.

4. acquiring time-series data relating to the movement of the second hand obtained by an inertial sensor attached to the second hand; The symptom information is output by inputting the time series data into the learning model. The program according to claim 1.

5. acquiring first time-series data relating to a movement of the first hand obtained by a first inertial sensor attached to the first hand; acquiring second time-series data relating to a movement of the second hand obtained by a second inertial sensor attached to the second hand; The symptom information is output by inputting the first time series data and the second time series data into the learning model. The program according to claim 1.

6. the symptom information indicates a probability of self-removal; If the probability is equal to or greater than a predetermined threshold, an alert is output. The program according to any one of claims 1 to 5.

7. the symptom information indicates a probability of self-removal; If the probability is equal to or greater than a predetermined threshold and the distance between the first hand and the second hand is within a predetermined range, an alert is output. The program according to claim 4 or 5.

8. If the probability is equal to or greater than a predetermined threshold, a command is output to the patient to provide a stimulus to prevent self-removal. The program according to claim 6.

9. If self-removal is detected, a second alert is output. The program according to claim 6.

10. Change the threshold or alert depending on the patient's condition The program according to claim 6.

11. time-series data relating to the movement of a second hand different from the first hand being punctured is acquired; The learning model outputs symptom information regarding signs of self-withdrawal of IV drips when time series data is input. By inputting the acquired time series data, symptom information is output. Information processing methods.

12. An information processing device having a control unit, The control unit time-series data relating to the movement of a second hand different from the first hand being punctured is acquired; The learning model outputs symptom information regarding signs of self-withdrawal of IV drips when time series data is input. By inputting the acquired time series data, symptom information is output. Information processing device.

13. a sensor that acquires time-series data regarding the movement of a second hand that is different from the first hand being punctured; an information processing device that inputs the acquired time series data into a learning model that outputs symptom information regarding symptoms of self-removal of an IV drip when the time series data is input, and outputs the symptom information; Equipped with Self-removal prevention system.

Images