JP2015103042A - Fall risk calculation system and notification system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a fall risk calculation system and a notification system capable of preventing falling out of a bed from occurring.SOLUTION: A fall risk calculation system 10 includes: a motion sensor 11 acquiring attitude information on a patient A; a server 20 calculating a fall risk of the patient A on the basis of the acquired attitude information and a risk calculation model for calculating the fall risk; and a terminal device 40 outputting the fall risk. The fall risk calculation system 10 configured in this way can monitor the risk and determine whether to go over to the patient A, so that it is possible to prevent the patient A from falling.

Description

  The present invention relates to a fall risk calculation system for calculating a risk of a subject falling from a bed and a notification system including the fall risk calculation system.

  In the medical field, an accident in which an inpatient falls from a bed is a big problem. As measures to prevent such a fall accident from the bed, a fall / fall assessment score is used in each hospital. This assessment score uses a score sheet that describes various items related to dysfunction, cognitive ability, etc., and the nurse can check the items for the patient to know the risk of the patient falling or falling. That's it.

  By using this assessment score to recognize the risk of falls and falls for each patient, nurses can pay more attention to high-risk patients by placing them around the bed and taking actions during hospitalization. it can.

  However, even if the assessment score indicates the risk of falls / falls for each patient in advance, nurses are not always paying attention to the high-risk patients. The fall cannot be completely prevented.

  As a means for solving such a problem, a notification device described in Patent Document 1 is known. In the notification device of Patent Document 1, a sensor is installed near the bed, and notification is performed when a patient is detected by the sensor. According to this reporting device, since it is possible to report a patient change, it is not necessary for the nurse to always be on the patient side.

  Similarly, a bed monitor device described in Patent Document 2 is also known. The bed monitor device of Patent Literature 2 arranges a load sensor on a bed, determines whether or not the patient's posture has become a predetermined posture from the load distribution of the load sensor, and the patient's posture becomes a predetermined posture. When it becomes, report is made. Even with this bed monitor device, the nurse does not always have to be on the patient side.

  In addition, a patient abnormality notification system described in Patent Document 3 is also known. The patient abnormality notification system of Patent Literature 3 is a system that performs notification when an abnormal state is detected, such as when a patient falls from a bed.

JP 2006-263396 A JP 2001-258957 A JP 2008-47097 A

  Since the reporting device of Patent Document 1 only detects the movement of the patient by the sensor, it is not known whether the behavior of the patient at that time is a really dangerous behavior, and thus erroneous detection may occur.

  Moreover, since the bed monitor apparatus of patent document 2 determines based on whether a patient's attitude | position is a predetermined attitude | position, naturally notification will not be performed if a patient's attitude | position does not become a predetermined attitude | position. Therefore, when the criterion for determination is set sweetly, the notification is made despite the low possibility of falling. On the other hand, if the criteria for judgment are set strictly, notification will not be made unless the possibility of falling is high, so when the nurse arrives after reporting, the patient may have already fallen from the bed It will occur.

  Furthermore, since the patient abnormality notification system of Patent Document 3 determines and notifies that the patient has fallen from the bed, it does not prevent the fall from the bed.

  The inventor of the present application has made various studies in order to solve such problems of the conventional technique, and as a result, calculates the risk of falling from the patient's posture information and outputs this risk. The present inventors have conceived that various problems can be solved and have completed the present invention.

  That is, an object of the present invention is to solve the above-described problems, and an object thereof is to provide a fall risk degree calculation system and a notification system that can be prevented before a fall accident from a bed.

  In order to solve the above problems, a fall risk level calculation system according to the present invention includes posture information acquisition means for acquiring posture information of a target person, the posture information acquired by the posture information acquisition means, and the fall of the target person. A risk calculating means for calculating the risk of falling of the subject by a risk calculating model for calculating a risk; an output means for outputting the risk of falling of the target calculated by the risk calculating means; Is provided.

Further, the fall risk level calculation system of the present invention is a plurality of calculation models in which the risk level calculation model is constructed for each score of the assessment score, and the fall risk level calculation system includes:
Assessment score information acquisition means for acquiring assessment score information of the subject;
Calculation model selection means for selecting the risk calculation model corresponding to the assessment score acquired by the assessment score information acquisition means, wherein the risk calculation means is the risk selected by the calculation model selection means The degree of fall risk of the subject is calculated using a degree calculation model.

  The fall risk calculation system according to the present invention is characterized in that the risk calculation model is a calculation model using fuzzy inference.

  Further, the fall risk calculation system of the present invention is a fuzzy inference adjustment unit that adjusts the parameters of the calculation model using the fuzzy inference using teacher data, a teacher data acquisition unit that acquires the teacher data, And the teacher data is a teacher risk level given to each posture with respect to a predetermined assessment score.

  The fall risk calculation system according to the present invention includes a fuzzy inference adjustment unit that adjusts parameters of a calculation model using the fuzzy inference using pseudo teacher data, and a pseudo teacher data generation unit that generates the pseudo teacher data. The pseudo-teacher data comprises a pseudo-risk calculated with respect to a predetermined posture using an assessment score and a pseudo-risk calculation formula, and the pseudo-risk calculation formula has a predetermined assessment score The coefficient is adjusted using learning data given for each posture.

  The fall risk calculation system according to the present invention is characterized in that an assessment score used for calculating the pseudo risk is acquired by the assessment score information acquisition means.

  The fall risk calculation system according to the present invention is characterized in that adjustment of parameters of a calculation model using the fuzzy inference is performed by a genetic algorithm.

  The fall risk calculation system according to the present invention is characterized in that adjustment of parameters of a calculation model using the fuzzy inference is performed by a fuzzy neural network.

  In the fall risk calculation system according to the present invention, the posture information of the subject acquired by the posture information acquisition unit is acquired by a motion sensor attached to the subject.

  Further, the notification system of the present invention is a notification system comprising any of the above-described fall risk calculation systems, report determination means, and notification means, wherein the notification determination means includes the fall risk calculation system. It is determined whether or not to make a notification based on the output from the output means, and the notification means makes a determination based on a determination result of the notification determination means.

  According to the fall risk calculation system of the present invention, since it is output as the fall risk of the subject based on the posture information of the subject, whether the risk should be monitored and whether or not to go to the subject Therefore, it is possible to prevent the subject from falling. Moreover, according to the notification system of the present invention, it is possible to realize a highly reliable notification system that prevents the subject from falling.

It is the block diagram which showed the structure of the fall risk degree calculation system of this invention. It is the figure which showed the example of application of the fall risk degree calculation system of this invention. It is a functional block diagram of the fall risk degree calculation system in the first embodiment. It is a flowchart of the fall risk degree calculation system in Embodiment 1. FIG. 5A is a front view of the patient sleeping on the bed, and FIG. 5B is a side view of the patient sitting on the bed. 6A is a front view of a patient sitting on the edge of the bed, and FIG. 6B is a side view of a patient sitting on the edge of the bed. It is a fuzzy rule in the risk calculation model of Embodiment 1. It is a membership function in the risk calculation model of Embodiment 1. It is a functional block diagram of the fall risk degree calculation system in the second embodiment. It is a flowchart of the fall risk degree calculation system in Embodiment 2. It is an example of the score sheet used for an assessment score. It is a functional block diagram of a fall risk degree calculation system in the third embodiment. 10 is a flowchart of a fall risk degree calculation system according to a third embodiment. This is an assessment score created virtually. It is a predetermined posture of postures 1-16. It is a teacher risk in a predetermined posture of postures 1-16. It is a figure which shows the X-axis acceleration membership function and the change location of each function. FIG. 18A is a membership function in the posture on the bed constructed by the real value GA, and FIG. 18B is a membership function in a state of sitting on the bed edge constructed by the real value GA. . It is the estimated risk calculated by the calculation model using the membership function of FIG. It is a structural diagram of a fuzzy neural network. This is a calculated value of a parameter adjusted by a fuzzy neural network. FIG. 22A is a membership function in the posture on the bed constructed by the fuzzy neural network, and FIG. 22B is a membership function in a state of sitting on the bed edge constructed by the fuzzy neural network. . It is the estimated risk calculated by the calculation model using the membership function of FIG. It is a functional block diagram of the fall risk degree calculation system in the fourth embodiment. It is a flowchart of the fall risk degree calculation system in Embodiment 4. This is the assumed data that has been set. It is an example of the calculated change rate. It is another example of the calculated change rate. It is the conceptual diagram performed by the pseudo risk degree calculation formula production | generation means. It is the block diagram which showed the structure of the notification system of this invention.

  Hereinafter, specific examples of the present invention will be described in detail with reference to the drawings. However, the embodiment shown below exemplifies a fall risk calculation system and a notification system for embodying the technical idea of the present invention, and the present invention is applied to the fall risk calculation system and the notification system. It is not intended to be specific. The present invention is equally applicable to a fall risk calculation system and a notification system according to other embodiments included in the claims.

  First, a basic configuration example of the fall risk degree calculation system 10 will be described with reference to FIG. FIG. 1 is a block diagram showing a configuration of the fall risk degree calculation system 10. The fall risk calculation system 1 is a system that is used for, for example, patients who are hospitalized in hospitals, elderly people living in elderly care facilities, elderly people who are cared at home, and the like. The risk of falling from the bed is calculated based on the posture information of the subject.

  Specifically, the fall risk calculation system 10 includes a motion sensor 11, a control unit 12, a storage unit 13, an input unit 14, and an output unit 15.

  The motion sensor 11 acquires the posture when the subject is on the bed and the posture when the subject is sitting on the edge of the bed as posture information of the subject. A specific example of the motion sensor 11 will be described later.

  The control unit 12 includes a CPU, a RAM, a ROM, and the like. The control unit 12 calculates the risk of falling of the subject from the posture information of the subject acquired by the motion sensor 11 based on a risk calculation model described later. Various processes in the calculation system 1 are performed based on a program stored in the ROM.

  The storage unit 13 is a storage device such as a dynamic random access memory (DRAM) or a hard disk drive (HDD), and stores various data temporarily and for a long time.

  The input unit 14 includes an input device such as a keyboard and a touch panel, and is used for inputting a check result of the assessment score of the target person described later.

  The output unit 15 includes a display device, a speaker, and the like, and outputs a result such as a fall risk degree calculated by the control unit 12.

  In addition, although the fall risk degree calculation system 10 uses the motion sensor 11 for acquiring the posture information of the subject person, the posture information of the subject person may be obtained by a method other than the motion sensor 11. Specifically, the fall risk level calculation system 10 may acquire the posture information by photographing the subject using a camera and extracting the posture information of the subject from the image. Further, the fall risk calculation system 10 places a mat equipped with a load sensor on the bed used by the subject, estimates posture information of the subject from the output of the load sensor, and acquires posture information. It doesn't matter. However, in the case of using a camera, the subject may not be photographed by the futon, and in the case of a load sensor, there is a problem that it is difficult to determine the posture. On the other hand, since the motion sensor 11 does not have such a worry and can be easily attached, the motion sensor 11 is preferable for obtaining posture information.

  Next, a specific application example of the fall risk degree calculation system 10 will be described with reference to FIG. FIG. 2 is a diagram showing an application example of the fall risk degree calculation system 10 for hospitalized patients. In this application example, the fall risk calculation system 10 calculates the fall risk of the patient A on the bed in the hospital room, and outputs the calculation result to the terminal device owned by the nurse.

  A motion sensor 11 is attached to the patient A. The motion sensor 11 is an IMU-Z manufactured by ZMP. The motion sensor 11 has acceleration, angular velocity, and geomagnetism sensors, and can measure three axes of XYZ. Further, three motion sensors 11 are attached to the patient A, one attached to the chest and one to the ankle. The motion sensor 11 includes a communication unit, and transmits the measured data as posture information of the patient A to the server 20 by wireless communication.

  The server 20 is a computer installed in a nurse station in a hospital ward. The server 20 includes the control unit 12, the storage unit 13, and the output unit 15 of the fall risk level calculation system 10. When the server 20 receives the posture information of the patient A transmitted from the motion sensor 11, the stored risk level is stored. Using the calculation model, the fall risk is calculated.

  The server 20 is connected to the network 30 in the hospital, and similarly outputs the fall risk level to the terminal device 40 possessed by a nurse connected to the network 30.

  The terminal device 40 is, for example, a portable small terminal, and includes a display unit 41 made of a liquid crystal panel, an input unit 42 made of a touch panel, and an audio output unit 43 made of a speaker. The terminal device 40 constitutes the output unit 15 of the fall risk degree calculation system 10, and the display unit 41 and the voice output unit 43 indicate the fall risk level of the patient A sent from the server 20 via the network 30. The information is output to the registered nurses. The terminal device 40 also constitutes the input unit 14 of the fall risk degree calculation system 10. The terminal device 40 inputs the check result of the assessment score of the patient A via the input unit 42, and sends it to the server 20 via the network 30. Can send.

[Embodiment 1]
Next, an embodiment of the fall risk degree calculation system will be described in detail with reference to FIGS. FIG. 3 is a block diagram illustrating functions of the fall risk degree calculation system 100 according to the first embodiment. FIG. 4 is a flowchart of the risk degree calculation system 100 according to the first embodiment. This fall risk degree calculation system 100 includes posture information acquisition means 110, risk degree calculation means 120, and output means 130.

  Specifically, the posture information acquisition unit 110 includes the motion sensor 11 described with reference to FIGS. 1 and 2. One motion sensor 11 is attached to the chest of the patient A and one to the ankle. The posture information acquisition unit 110 acquires posture information of the patient A by measuring the accelerations of the X axis, the Y axis, and the Z axis with the motion sensor 11. In the present embodiment, the posture information acquisition unit 110 measures the static acceleration using the motion sensor 11. The posture information acquisition unit 110 performs measurement every 0.1 s using the motion sensor 11, and transmits posture information to the server 20 via the communication unit of the motion sensor 11 each time.

  Specifically, the risk degree calculation unit 120 includes the control unit 12 and the storage unit 13 described in FIG. 1 and the server 20 in FIG. When the posture information acquisition unit 110 acquires posture information (step 1101), the risk level calculation unit 120 stores in advance the risk level of the posture of the patient A in the storage unit 13 based on the acquired posture information. The risk level calculation model is used (step 1102). Although a specific example of this risk level calculation model will be described later, any calculation model may be used as long as the risk level can be calculated based on the posture information of the patient A who is the subject.

  Specifically, the output unit 130 includes the output unit 15 described in FIG. 1 and the terminal device 40 in FIG. When the risk level of the patient A is calculated by the risk level calculation unit 120, the output unit 130 displays the risk level on the display unit 41 of the terminal device 40 (step 1103).

  The risk level output to the display unit 41 by the output unit 130 is, for example, a percentage display such as a risk level of 80% and a risk level of 30%, a level display such as a high risk level and a low risk level, It can be represented by different colors depending on the degree. In addition, the voice output unit 43 can be used to express the voice.

  As described above, the fall risk calculation system 100 calculates the risk based on the posture information of the patient A, instead of notifying only by detection by the sensor as in the conventional example shown in Patent Document 1. Therefore, there is no false alarm due to false detection.

  Further, unlike the conventional example shown in Patent Document 2, the fall risk calculation system 100 does not simply determine whether or not the patient's posture is a predetermined posture and report it. It is possible to determine whether or not to go to the patient A by his / her own judgment by looking at the risk level of the patient A displayed on the display unit 41 of 40. Therefore, for example, the nurse determines the degree of risk that the patient A should go to the patient A in comparison with the assessment score result of the patient A that is known in advance. You can also

  Here, an example of the risk calculation model for calculating the risk of the patient A will be specifically described. First, in the present embodiment, the risk calculation model for calculating the risk based on the posture information of the patient A is a calculation model based on fuzzy inference. This calculation model based on fuzzy reasoning is composed of fuzzy rules and membership functions defined by the static acceleration and risk of the patient A.

  In the present embodiment, the fuzzy rules and the membership functions used are different between the posture of the patient A on the bed and the posture of the patient A sitting on the edge of the bed.

  Next, conditions when the fuzzy rules and membership functions are defined will be described with reference to FIGS. FIG. 5A is a front view of the patient A sleeping on the bed b, and FIG. 5B is a side view of the patient A sitting on the bed b. FIG. 6A is a front view of the patient A sitting on the edge of the bed b, and FIG. 6B is a side view of the patient A sitting on the edge of the bed b.

  This time, as shown in FIG. 5A, a fence (or wall) c is arranged on three sides of the bed b used by the patient A. Patient A is in a situation where there is no fence c on the left side as viewed from patient A.

  As shown in FIG. 5, when the patient A is on the bed b, the static acceleration in the X-axis direction and the Y-axis direction of the motion sensor 11 on the chest is used as posture information. Regarding the X-axis direction, the static acceleration approaches -1G as the patient A faces the left side (side without the fence c), and the static acceleration approaches 1G as the patient A faces the right side (side with the fence c). . Regarding the Y-axis direction, the static acceleration approaches 1G as the patient A in FIG. 5A lies down to the state where the upper body in FIG.

  As shown in FIG. 6, when the patient A is sitting on the edge of the bed b, the static acceleration in the X-axis direction and the Z-axis direction of the motion sensor 11 on the chest is used as posture information. With respect to the X-axis direction, the static acceleration approaches 1G as the upper body leans to the right when viewed from the patient A, and the static acceleration approaches -1G as the upper body leans to the left when viewed from the patient A. Regarding the Z axis, the static acceleration approaches 1G as the upper body leans backward from the patient A, and the static acceleration approaches -1G as the upper body leans forward from the patient A.

  Whether the patient A is sitting on the bed b or the edge of the bed b was determined using the ankle motion sensor 11. Specifically, for example, when the Y-axis static acceleration is 0 G in either one of the motion sensors 11 of both ankles, the posture of the patient A is considered as the posture on the bed b, and the Y-axis static When both accelerations are 1G, the posture of the patient A is distinguished as a posture in a state where the patient A is placed on the edge of the bed b.

  FIG. 7 shows the fuzzy rule used this time under the above conditions, and FIG. 8 shows the membership function. In addition, regarding the membership function of FIG. 8, the membership function in the X-axis acceleration in a state where the patient A is sitting on the edge of the bed b is shown.

  As described above, as a risk calculation model for calculating the risk level of the patient A, as an advantage of a calculation model using fuzzy reasoning, empirical knowledge such as nurse's know-how can be described as a rule. This is because it is suitable for modeling “skills”.

[Embodiment 2]
Next, another embodiment of the fall risk degree calculation system will be described with reference to FIGS. 9 and 10. FIG. 9 is a block diagram illustrating functions of the fall risk degree calculation system 200 according to the second embodiment. FIG. 10 is a flowchart of the degree-of-risk calculation system 200 according to the second embodiment. The fall risk calculation system 200 includes posture information acquisition means 210, risk calculation means 220, output means 230, assessment score information acquisition means 240, fuzzy inference model selection means 250, and fuzzy inference model storage means 260.

  By the way, currently, in the medical field, when a patient is hospitalized, it is obliged to create a “fall / fall assessment score” for the purpose of investigating the possibility of the patient's fall / fall. The “fall / fall assessment score” is performed using a score sheet on which various items relating to functional impairment and cognitive ability are described. By using this score sheet to check and evaluate items for each patient, the nurse pays attention to the arrangement around the bed and behavior during hospitalization in consideration of the possibility of the patient falling or falling. In other words, the nurse determines whether or not the patient's posture is a dangerous posture based on the “fall / fall assessment” indicating the patient's fall / fall risk. By utilizing such “skills” of nurses in the fall risk calculation system, it becomes possible to calculate the fall risk of patients with higher reliability. Therefore, the fall risk level calculation system 200 according to the second embodiment is a system that uses the fall / fall assessment score utilizing the “skill” of the nurse.

  The posture information acquisition unit 210, the risk level calculation unit 220, and the output unit 230 are the same as the posture information acquisition unit 110, the risk level calculation unit 120, and the output unit 130 in the risk level calculation system 100 of the first embodiment. Also in this embodiment, the risk calculation model is a calculation model based on fuzzy inference.

  The assessment score information acquisition unit 240 includes the input unit 14 described with reference to FIG. 1 and the display unit 41 and the input unit 42 of the terminal device 40 described with reference to FIG. This assessment score information acquisition means 240 acquires the result of the assessment score confirmed for every patient using the score sheet | seat in which various items regarding a functional disorder, cognitive ability, etc. are described. Specifically, a score sheet is displayed on the display unit 41 of the terminal device 40, and the nurse confirms each item via the input unit 42, and transmits this entry result to the server 20, so that each patient is confirmed. Obtain assessment score information (step 2101)

  FIG. 11 is an example of a score sheet used for the assessment score. A score sheet composed of such items is displayed on the display unit 41 of the terminal device 40. When the nurse checks the corresponding item of each evaluation content via the input unit 42, the score sheet is automatically set based on the evaluation value. A total score is calculated. Then, the result including the check result for each item is transmitted to the server 20. In addition, the score sheet is not unified at the present time, and varies depending on each hospital. Accordingly, there are various kinds of evaluation items such as more score sheets and fewer score sheets than the score sheet of FIG. 11, and there are many types of score sheets.

  The fuzzy inference model selection unit 250 includes the control unit 12 described with reference to FIG. 1 and the server 20 described with reference to FIG. The fuzzy inference model storage unit 260 includes the storage unit 13 described with reference to FIG. 1 and the server 20 described with reference to FIG.

  The calculation model using fuzzy reasoning described in the first embodiment calculates the risk level of a patient using one fuzzy rule and a membership function that are not related to the assessment score entry result. However, the nurse derives the degree of risk from the “patient posture” based on the patient's “fall / fall assessment score” in an empirical and empirical manner. In other words, the assessment score and the patient's posture seem to be greatly related to the judgment of the risk of falling unconsciously performed by the nurse.

  Therefore, in the fall risk calculation system 200, in order to approach such a judgment of the nurse, the fuzzy inference model storage unit 260 has a calculation model corresponding to each pattern of the result of the assessment score shown in FIG. It is remembered.

  Then, the fuzzy inference model selection unit 250 selects a calculation model having the same pattern as the pattern of the result of the assessment score acquired by the assessment score information acquisition unit 240 from the fuzzy inference model storage unit 260 (Step 2102). Note that the calculation model selected here is temporarily stored in a RAM or the like constituting the storage unit 13.

  When the posture information acquisition means 210 acquires patient posture information (step 2103), the risk level is calculated using the calculation model selected in step 2102 (step 2104). The degree is output (step 2105).

  As described above, the fall risk calculation system 200 according to the second embodiment stores the calculation model corresponding to the pattern of the assessment score result, selects the calculation model having the same pattern as the written result, and selects the risk model. Therefore, it is possible to get closer to the judgment made by the nurse.

[Embodiment 3]
Next, another embodiment of the fall risk degree calculation system will be described with reference to FIGS. FIG. 12 is a block diagram illustrating functions of the fall risk degree calculation system 300 according to the third embodiment. FIG. 13 is a flowchart of the risk calculation system 300 according to the third embodiment. This fall risk degree calculation system 300 includes posture information acquisition means 310, risk degree calculation means 320, output means 330, assessment score information acquisition means 340, fuzzy inference model selection means 350, fuzzy inference model storage means 360, teacher data acquisition means. 370 and fuzzy inference model adjusting means 380.

  In the fall risk degree calculation system according to the first embodiment or the second embodiment, a description is given using a calculation model based on fuzzy reasoning as a risk degree calculation model when calculating a risk degree from patient posture information.

  In fuzzy inference, work that optimizes the inference results is performed by setting and adjusting fuzzy rules and membership functions. In the first embodiment and the second embodiment, such work may be performed by trial and error, and the fuzzy rules and membership functions after the adjustment may be used. In order to enable adjustment of inference parameters, a teacher data acquisition unit 370 and a fuzzy inference model adjustment unit 380 are provided.

  Posture information acquisition means 310, risk degree calculation means 320, output means 330, assessment score information acquisition means 340, fuzzy inference model selection means 350 and fuzzy inference model storage means 360 are the posture information in the risk degree calculation system 200 of the second embodiment. The acquisition unit 210, the risk level calculation unit 220, the output unit 230, the assessment score information acquisition unit 240, the fuzzy inference model selection unit 250, and the fuzzy inference model storage unit 260 are the same.

  The teacher data acquisition unit 370 includes the input unit 14 described with reference to FIG. 1, the display unit 41 of the terminal device 40 described with reference to FIG. 2, and the input unit 42. The teacher data acquisition unit 370 acquires teacher data used for adjusting the fuzzy inference model for each assessment score entry result performed by the fuzzy inference model adjustment unit 380.

  This teacher data is the risk (teacher risk) for each predetermined posture with respect to a certain assessment score. More specifically, the teacher data is assumed to be a patient with a virtually created assessment score, and the degree of risk of a predetermined posture in this patient is expressed as 0 to 100%. .

  An example of teacher data in the present embodiment is shown in FIG. FIG. 14 shows an assessment score created virtually. Assuming that the patient of the score sheet of FIG. 14 is on the bed, FIG. 16 shows the results of the risk (teacher risk) in each of the predetermined postures 1 to 16 shown in FIG. FIG. 16 shows a description of the state of each posture in postures 1 to 16 and each risk level. And such teacher data are acquired for every pattern of an assessment score.

  15 are postures on the bed, and postures 10 to 16 are postures while sitting on the bed edge. In the present embodiment, the postures 1 to 16 are set as the predetermined postures, but teacher data may be acquired assuming more postures. Moreover, although each risk degree in the predetermined posture shown in FIG. 16 is obtained from one nurse, it may be obtained from a plurality of nurses and each average value may be used as teacher data.

  The risk level calculation system 300 acquires such teacher data by the teacher data acquisition unit 370. Specifically, the teacher data acquisition unit 370 displays the virtually generated assessment score of FIG. 14 on the display unit 41 of the terminal device 40 and the predetermined postures 1 to 16 shown in FIG. The nurse inputs the degree of risk in the case via the input unit 42 and transmits the input result to the server 20 to obtain teacher data (step 3101).

  In addition, in this embodiment, although the example which acquired the teacher data using the terminal device 40 was shown, it is not limited to such an acquisition example. For example, the configuration may be such that the assessment score virtually created by the nurse and a predetermined posture are shown in advance, the degree of risk is entered, and the entry result is input using the input unit provided in the server 20.

  The fall risk calculation system 300 adjusts the fuzzy inference model in the fuzzy inference model adjustment unit 380 based on the teacher data acquired by the teacher data acquisition unit 370 (step 3102).

  The fuzzy inference model adjusting unit 380 is configured by the control unit 12 described in FIG. 1 and the server 20 described in FIG. A specific adjustment example of the fuzzy inference model will be described later. The fall risk calculation system 300 adjusts the fuzzy inference model for each assessment score using the teacher data, and stores the adjusted fuzzy inference model in the fuzzy inference model storage unit 360 (step 3103).

  When the assessment score information of the patient A is acquired by the assessment score information acquisition means 340 (step 3104), an inference having the same pattern as the assessment score of the patient A is selected from the inference models stored in the fuzzy inference model storage means 360. The model is selected by the fuzzy inference model selection means 350 (step 3105).

  When the posture information acquisition unit 310 acquires the posture information of the patient A (step 3106), the risk level calculation unit 320 uses the inference model selected by the fuzzy inference model selection unit 350 to determine the risk level of the patient A. The risk is calculated (step 3107), and the degree of risk is output by the output means 330 (step 3108).

  As described above, the fall risk calculation system 300 according to the present embodiment obtains teacher data for the calculation model using fuzzy reasoning and performs optimization adjustment, so that the nurse performs the calculation. You can be closer by judgment.

  Here, an example of adjusting the fuzzy inference model using the acquired teacher data will be specifically described. Since there are various methods for adjusting the parameters of the fuzzy inference model, any method may be used. In this embodiment, the fuzzy inference model is adjusted by a method using a real value GA (Genetic Algorithm) and a method using an FNN (fuzzy neural network).

  First, the real value GA will be described. A genetic algorithm is a learning algorithm that mimics the process by which a living organism adapts to the environment and evolves. In the evolutionary process of living organisms in nature, solids that adapt to the environment in the population of individuals survive with a high probability and leave children in the next generation. The concept of GA is to model this mechanism and find a solid that best adapts to the environment, that is, a solution that gives an optimum value for the objective function.

The procedure of processing using real value GA is as follows:
1. The teacher risk obtained from the teacher data was used as the GA evaluation function.
2. The parameters were changed by GA using the width and height of the membership function as parameters.
3. As an evaluation function, a membership function was created with parameters calculated from GA, and the risk level was calculated based on the attitude of the teacher data, and an error between the calculated risk level and the risk level set by the nurse was obtained.
4). The parameter with the highest evaluation (small error) when the sum of errors satisfies the termination condition or the maximum number of generations is reached was adopted.

  The parameters to be adjusted are, for example, the width and height of the straight lines of the three membership functions (X-axis acceleration, Z-axis acceleration, risk) if the patient is sitting on the edge of the bed. Number × width and height to be changed = 3 × 7 = 21 items. FIG. 17 is a diagram for explaining the X-axis acceleration membership function and the changed portions of each function. The membership function of the X-axis acceleration is composed of three functions (NE, ZE, PO) as shown in FIG. 17, and the width and height of these three functions are changed by a real value GA. Similarly, the Z-axis acceleration and the risk level were also changed.

  A membership function constructed by this real value GA is shown in FIG. FIG. 18A shows the membership function in the posture on the bed, and FIG. 18B shows the membership function in a state of sitting on the bed edge. Further, FIG. 19 shows the risk (estimated risk) calculated by the calculation model using the membership function shown in FIG. FIG. 19 shows not only the estimated risk, but also the error between the estimated risk and the teacher data of the postures 1 to 9 in the posture on the bed and the postures 10 to 16 sitting on the bed edge as the verification result. Show. As can be seen from FIG. 19, the error from the teacher data was 6% at the maximum, and it was confirmed that the learning accuracy by the real value GA was high.

  Next, a fuzzy neural network will be described. The fuzzy neural network combines "fuzzy inference with the feature that human ambiguity can be expressed linguistically by IF-THEN rule" and "neural network with the feature that any input / output relationship can be identified by the learning function" Method. As features, fuzzy rules can be identified and membership functions can be automatically adjusted by learning a neural network.

  FIG. 20 is a structural diagram of the fuzzy neural network constructed this time. Learning was performed by giving acceleration (posture) as input (x1, x2) and teacher risk as output (y *). A sigmoid function was used as the function. In the fuzzy neural network, the membership function is adjusted and the fuzzy rule is identified by adjusting the three kinds of weights ωc, ωg, and ωb by the error back propagation method (BP method).

  FIG. 21 shows calculated values of ωc, ωg, and ωb adjusted by this fuzzy neural network. FIG. 22 shows a part of the membership function constructed by the fuzzy neural network. FIG. 22A shows the X-axis acceleration used in the posture on the bed and the membership function in the Y-axis acceleration, and FIG. 22B shows the X-axis acceleration used when sitting on the bed edge and the Z-axis. Membership function related to acceleration. FIG. 23 shows the risk (estimated risk) calculated by the calculation model adjusted by the fuzzy neural network. FIG. 23 shows not only the estimated risk but also the error between the teacher data of postures 1 to 16 and the estimated risk as a verification result. As can be seen from FIG. 23, the error from the teacher data used was 13% at the maximum, which was larger than the real value GA, but was in an allowable range.

  From the above, it can be seen that if the parameter adjustment of the fuzzy inference model is performed with the real value GA, the degree of freedom of adjustment of the membership function is high, and thus the model accuracy is high. In addition, when performed by a fuzzy neural network, the model accuracy is somewhat inferior to that of the real value GA, but it can be determined based on teacher data including fuzzy rules.

[Embodiment 4]
Next, another embodiment of the fall risk degree calculation system will be described with reference to FIGS. FIG. 24 is a block diagram illustrating functions of the fall risk degree calculation system 400 according to the fourth embodiment. FIG. 25 is a flowchart of the risk degree calculation system 400 according to the fourth embodiment. The fall risk calculation system 400 includes posture information acquisition means 410, risk calculation means 420, output means 430, assessment score information acquisition means 440, fuzzy inference model selection means 450, fuzzy inference model storage means 460, fuzzy inference model adjustment. Means 470 and pseudo teacher data generation means 480 are provided.

  Here, in the fall risk degree calculation system 300 of the third embodiment, the teacher data acquisition unit 370 acquires teacher data when adjusting the fuzzy inference model. The teacher data is a risk level (teacher risk level) for each predetermined posture with respect to the assessment score result. Therefore, if the assessment score results are different, the teacher risk in a predetermined posture is also different. In this regard, in the third embodiment, all the teacher data is acquired for each pattern of the result of the assessment score. The teacher data is input by the nurse via the input unit 42 of the terminal device 40.

  However, although an example of the score sheet is shown in FIG. 11, the number of items of the score sheet varies depending on each hospital. In the case of a score sheet with a small number of items, it is possible to obtain teacher data using all patterns of entry results in this score sheet, but 49 items (age of evaluation items) as in the score sheet of FIG. To the sitting position balance), it is difficult to input by the nurse because there are so many patterns that the teacher data is acquired using all the patterns of the assessment score. Note that the realistic number that can be input by the nurse is considered to be about 100 patterns.

  Therefore, the fall risk calculation system 400 according to the fourth embodiment includes a pseudo teacher data generation unit 480. When the fuzzy inference model adjustment unit 470 adjusts the fuzzy inference model, the pseudo teacher data generation unit 480 uses the teacher data generation unit 480 as teacher data. By using pseudo teacher data generated in a pseudo manner, acquisition of teacher data is simplified.

  First, posture information acquisition means 410, risk level calculation means 420, output means 430, assessment score information acquisition means 440, fuzzy inference model selection means 450, fuzzy inference model storage means 460 and fuzzy inference model adjustment means 470 are the same as those in the third embodiment. Attitude information acquisition means 310, risk calculation means 320, output means 330, assessment score information acquisition means 340, fuzzy inference model selection means 350, fuzzy inference model storage means 360, and fuzzy inference model adjustment means 380 It is the same.

  The pseudo-teacher data generation unit 480 includes a pseudo-risk level calculation formula creating unit 481 that creates a pseudo-risk level calculation formula for generating pseudo-teacher data, and learning data acquisition necessary for creating the pseudo-risk level calculation formula Means 482 and assumption data generation means 483 are provided.

  The point of generating pseudo teacher data by the pseudo teacher data generation means 480 will be described.

  First, a pseudo risk level calculation formula for generating pseudo teacher data is created by the pseudo risk level formula creation means 481.

  First, a gene is defined in which the number of items in the score sheet is the number of elements (49 in the case of the score sheet in FIG. 11), and the content of each element is the rate of change.

The default value D in equation (1) is the minimum risk level (risk level when there is no check in the score sheet) in each posture, and the rate of change of the checked item is added to D. Thus, the pseudo risk y ^* _p is calculated. Regarding the setting of the default value D, for example, it is conceivable that the nurse performs a risk evaluation when the check sheet is not checked, and the evaluation value is set as the default value D.

  Next, learning data is acquired by the learning data acquisition means 482. This learning data is the degree of risk evaluated by the nurse, and is the degree of risk of each posture 1 to 16 in an assessment score of a certain pattern. As the learning data, in this embodiment, 100 patterns of assessment scores are used to evaluate the risk of postures 1 to 16 by a nurse. The specific learning data is acquired by using the terminal device 40 or the server 20 as in the teacher data acquisition unit 370 of the third embodiment. FIG. 26 shows a part of the learning data acquired by the learning data acquisition means 482.

Pseudo risk calculation formula creation unit 481, the learning and true value y _p the risk of learning data acquired by the data acquisition unit 482, the following average error Err is minimized change rate (of formula (2) ( The pseudo risk level calculation formula is created by finding the combination of 1) μ _i ) with the real value GA.

  In the present embodiment, the real value GA is set by using an elite strategy for the generation change model, the SPX method for the crossover method, 20000 for the maximum generation, 1000 for the number of individuals, 49 for the number of dimensions, and an end condition of formula (2 ) Average error ≦ 0.5%, the number of assumed data is 100, and the initial random number range is 0.0-20.0. The pseudo risk level calculation formula creating means 481 creates the formula (1) for each posture of postures 1 to 16 by calculating the change rate of each item by the real value GA.

  An example of the calculated change rate is shown in FIGS. FIG. 27 shows the rate of change of posture 8 shown in FIG. 15 (inclined to the side descending on the bed (small)), and FIG. 28 shows posture 12 shown in FIG. )) Change rate. At this time, the default value of the formula (1) was D = 0.

  Next, hypothetical data is acquired by the hypothetical data generation means 483. This assumption data is an assessment score composed of an unknown check pattern that is not used when learning data is acquired by the learning data acquisition means 482, and is created by the server 20 or the like. The pseudo teacher data generation unit 480 can calculate the pseudo risk by inputting the unknown assessment score into the pseudo risk calculation formula created by the pseudo risk calculation formula creation unit 481. Therefore, the assessment score can be calculated. Thus, pseudo-teacher data covering all the patterns is generated. The average error (Err obtained by equation (2)) between the pseudo risk obtained from the learning data used to calculate the rate of change and the true value of the risk is 0.93%, and the maximum error is 3.26%.

  As described above, using the learning data acquired by the learning data acquisition unit 482, the pseudo risk level calculation formula creating unit 481 calculates the change rate of the formula (1) that is the pseudo risk level calculation formula, and the formula (1). Create The pseudo risk level calculation formula creating means 481 creates such a risk level formula for each of the postures 1 to 16 in FIG. FIG. 29 shows the concept performed by the pseudo risk level calculation formula generation means 481.

  Then, as shown in FIG. 29, it becomes possible to calculate the degree of risk based on the difference in the assessment score for each posture, and the pseudo teacher data generation unit 480 uses the formula created by the pseudo risk level calculation formula creation unit 481. Then, using the assessment score composed of an unknown check pattern that is the assumption data generated by the assumption data generation means 483, pseudo teacher data that covers all the patterns of the assessment score is generated.

  After that, as in the third embodiment, the fuzzy inference model adjusting unit 470 adjusts the fuzzy inference model using the pseudo teacher data generated by the pseudo teacher data generating unit 480 (step 4101). The adjusted fuzzy inference model is stored in the fuzzy inference model storage means (step 4102). When the assessment score information of the patient A is acquired by the assessment score information acquisition means 440 (step 4103), the assessment score information is the same as the assessment score of the patient A out of the inference models stored in the fuzzy inference model storage means 460. The fuzzy inference model selection means 450 selects an inference model (step 4104).

  When the posture information acquisition unit 410 acquires the posture information of the patient A (Step 4105), the risk level calculation unit 420 uses the inference model selected by the fuzzy inference model selection unit 450 to calculate the risk level of the patient A. The risk is calculated (step 4106), and the degree of risk is output by the output means 430 (step 4107).

  In the present embodiment shown in FIG. 24, pseudo-teacher data corresponding to the assessment scores of all patterns is generated in advance, and the fuzzy inference model is adjusted for each assessment score (fuzzy inference model adjusting means 470). This is a configuration for storing the performed fuzzy inference model (fuzzy inference model storage means 460). In such a configuration, when the assessment score information of the patient A is acquired by the assessment score information acquisition unit 440, the fuzzy inference model selection unit 450 can easily select the fuzzy inference model, but the fuzzy inference model storage unit 460 The amount of data stored in advance will increase. Therefore, when the assessment score information of the patient A is acquired by the assessment score information acquisition unit 440, the fall risk calculation system generates only the pseudo teacher data of the assessment score corresponding to the patient A by the pseudo teacher data generation unit 480. However, the configuration may be such that only the pseudo-teacher data is used to adjust the fuzzy inference model of the assessment score corresponding to the patient A. With such a configuration, the fall risk calculation system can reduce the amount of data stored in advance.

  Here, in this embodiment, the pseudo risk degree calculation formula creating means 481 calculates the optimum solution of the change rate of the formula (1) by the real value GA. By finding this optimal solution, it was possible to visualize which items of the assessment score are important for each posture. That is, a high rate of change means that the degree of importance in evaluation is high, and it can be seen that this is an item that is important in the judgment of the nurse. This can be used, for example, to instruct which items on the score sheet should be emphasized in an educational situation for a new nurse. Further, by extracting more highly important items from the score sheet having many items as shown in FIG. 11, it can be used to create a simplified score sheet. Although the real value GA is used as a method for obtaining the optimum solution, it may be obtained by a brute force method.

  Next, an embodiment of a notification system realized by using the fall risk degree calculation system of Embodiments 1 to 4 will be described with reference to FIG. FIG. 30 is a block diagram illustrating functions of the notification system 500 in the present embodiment.

  The notification system 500 includes a fall risk degree calculation system 510, a notification determination unit 520, and a notification unit 530. The notification system 500 can also be realized by using the fall risk system shown in FIG. Therefore, description will be made with reference to FIG. 2 together with FIG.

  The fall risk calculation system 510 can use any of the fall risk calculation systems shown in the first to fourth embodiments.

  When the risk level data output from the output unit constituting the fall risk level calculation system 510 is input, the report determination unit 520 determines whether or not the risk level exceeds a preset risk level. I do. Then, the notification determination unit 520 outputs a notification instruction signal when it is determined that the preset risk level is exceeded. In addition, the report determination means 520 is comprised by the server 20 of FIG. 2, for example.

  The reporting unit 530 includes a reporting unit. When the reporting unit 530 receives the reporting instruction signal output from the reporting determination unit 520, the reporting unit 530 reports using the reporting unit. Specifically, the notification means 530 is provided in the voice output unit 43 of the terminal device 40 in FIG. 2, the notification device provided in the nurse station where the server 20 is installed, the hospital room used by the patient A, and its surroundings. It can be configured with the arranged reporting device.

  When the notification is made by the reporting means 530, the nurse can know that the patient A has a risk of falling, so that the nurse can rush to the patient A before falling. .

  As described above, according to the fall risk level calculation system in the present embodiment, the risk level of the subject person is output based on the posture information of the subject person. Therefore, it is possible to prevent the subject from falling down. Moreover, according to the notification system in the present embodiment, it is possible to realize a highly reliable notification system that prevents the subject from falling.

10, 100, 200, 300, 400, 510 ... Falling risk calculation system 11 ... Motion sensor 12 ... Control unit 13 ... Storage unit 14 ... Input unit 15 ... Output unit 20 ... Server 30 ... Network 40 ... Terminal device 41 ... Display Unit 42 ... input unit 43 ... audio output unit 110, 210, 310, 410 ... posture information acquisition means 120, 220, 320, 420 ... risk level calculation means 130, 230, 330, 430 ... output means 240, 340, 440 ... Assessment score information acquisition means 250, 350, 450 ... fuzzy inference model selection means 260, 360, 460 ... fuzzy inference model storage means 370 ... teacher data acquisition means 380, 470 ... fuzzy inference model adjustment means 480 ... pseudo teacher data generation means 481 ... Pseudo risk calculation formula creation means 482 ... Learning data acquisition means 83 ... assumed data generating means 500 ... report system 520 ... notification determination unit 530 ... Problem means

Claims (10)

Posture information acquisition means for acquiring posture information of the subject person;
A risk degree calculating means for calculating the risk of falling of the subject by the risk information calculating model for calculating the risk of falling of the subject and the posture information acquired by the posture information acquiring means;
A fall risk level calculation system comprising: output means for outputting the fall risk level of the subject calculated by the risk level calculation means. The risk calculation model is a plurality of calculation models constructed for each assessment score,
The fall risk calculation system is
Assessment score information acquisition means for acquiring assessment score information of the subject;
Calculation model selection means for selecting the risk calculation model corresponding to the assessment score acquired by the assessment score information acquisition means;
With
The fall risk degree calculation system according to claim 1, wherein the risk degree calculation means calculates the fall risk degree of the subject using the risk degree calculation model selected by the calculation model selection means. .   The fall risk calculation system according to claim 1 or 2, wherein the risk calculation model is a calculation model using fuzzy reasoning. The fall risk calculation system is
Fuzzy inference adjustment means for adjusting the parameters of the calculation model using the fuzzy inference using teacher data;
Teacher data acquisition means for acquiring the teacher data;
With
The fall risk calculation system according to claim 3, wherein the teacher data is a teacher risk assigned to each posture with respect to a predetermined assessment score. The fall risk calculation system is
Fuzzy inference adjustment means for adjusting the parameters of the calculation model using the fuzzy inference using pseudo-teacher data;
Pseudo teacher data generation means for generating the pseudo teacher data;
With
The pseudo-teacher data includes a pseudo risk calculated for a predetermined posture using an assessment score and a pseudo risk calculation formula,
4. The fall risk calculation system according to claim 3, wherein the pseudo risk calculation formula is such that a coefficient is adjusted using learning data given to each posture with respect to a predetermined assessment score. .   6. The fall risk degree calculation system according to claim 5, wherein the assessment score used for calculating the pseudo risk degree is obtained by the assessment score information obtaining means.   The fall risk calculation system according to any one of claims 4 to 6, wherein the adjustment of the parameters of the calculation model using the fuzzy inference is performed by a genetic algorithm.   The fall risk calculation system according to any one of claims 4 to 6, wherein the adjustment of the parameters of the calculation model using the fuzzy inference is performed by a fuzzy neural network.   The degree of risk of falling according to any one of claims 1 to 8, wherein the posture information of the subject acquired by the posture information acquisition means is acquired by a motion sensor attached to the subject. Calculation system. A fall risk calculation system according to any one of claims 1 to 9,
Report determination means;
Reporting means;
A reporting system comprising:
The notification determination means determines whether or not to report based on the output from the output means of the fall risk calculation system,
The reporting system, wherein the reporting unit makes a determination based on a determination result of the reporting determination unit.