Water balance management system, inference device, trained model generation device, trained model generation method, and computer program

Abstract

To improve the management property of information on moisture balance of a subject.SOLUTION: A detection device 11 outputs a detection signal DT corresponding to a load applied to a bed 20 on which a subject lies. An inference device 12 acquires load change information indicating a temporal change in the load on the basis of the detection signal DT, and infers an event that causes moisture balance of the subject from the load change information. An output device 13 outputs inference information PR corresponding to an inference result of the event.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a system for managing fluid balance information based on fluid intake and excretion of a subject. The present invention also relates to an inference device that infers events that cause the fluid balance of a subject, and a computer program executable by a processor mounted on the inference device. The present invention also relates to a generation device that generates a trained model used for inference by the inference device, and a computer program executable by a processor mounted on the generation device. The present invention also relates to a method for generating the trained model. [Background technology]

[0002] In medical facilities, the weight of a patient, which is an example of a subject, is measured once a day or once every few days. The fluid balance is managed by comparing the weight gain or loss caused by the fluid intake and urinary excretion by the patient during the period since the previous weight measurement with the weight change of the patient during the period. Patent Document 1 discloses a system for electronically managing fluid balance information. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2001-337998 A Summary of the Invention [Problem to be solved by the invention]

[0004] An object of the present invention is to improve the manageability of a subject's water balance information. [Means for solving the problem]

[0005] A first aspect of the present invention to achieve the above object is a water balance management system, A detection device that outputs a detection signal corresponding to a load applied to a bed on which a subject is lying; an inference device that acquires load change information indicating a change over time of the load based on the detection signal, and infers an event that causes water balance of the subject from the load change information; an output device that outputs inference information corresponding to an inference result of the event; It is equipped with:

[0006] A second aspect of the present invention for achieving the above object is an inference device, comprising: an interface that receives load change information indicating a change over time in a load applied to a bed on which a subject is lying, the change being acquired based on a detection signal corresponding to the load; A processor that infers an event that causes the subject's fluid balance from the load change information and outputs information corresponding to the inference result of the event to an output device; It is equipped with:

[0007] A third aspect for achieving the above object is a computer program executable by a processor installed in an inference device, comprising: When executed, the inference device: receiving load change information indicating a change over time in a load applied to a bed on which a subject is lying, the change being acquired based on a detection signal corresponding to the load; Inferring an event that causes fluid balance in the subject from the load change information; Information corresponding to the inference result of the event is output to an output device.

[0008] According to the configurations of the first to third aspects, the weight change associated with the fluid balance of the subject while getting out of bed is reflected in the change in the load applied to the bed when the subject returns to bed. Since the event causing the fluid balance is inferred based on the change, the management of the fluid balance during getting out of bed can be automated. This allows the fluid balance during getting out of bed, which was previously excluded from management, to be easily included in the management, improving the manageability of the fluid balance information of the subject not only while in bed but also while getting out of bed. In addition, the complementary recording of the fluid balance during getting out of bed, which was previously generally done manually, can be automated, reducing the burden on the user.

[0009] A fourth aspect for achieving the above object is a trained model generation device, comprising: an interface that receives, as learning data, a combination of load change information indicating a change over time in a load applied to a bed on which a subject is lying, the load change information being acquired based on a detection signal corresponding to the load, and event information indicating an event that causes fluid balance in the subject and is associated with the change over time; A processor that performs machine learning based on the learning data to generate a learned model that infers the event from the load change information; It is equipped with:

[0010] A fifth aspect for achieving the above object is a computer program executable by a processor mounted on a trained model generation device, By executing the above, the trained model generation device Accepting, as learning data, a combination of load change information indicating a change over time in a load applied to a bed on which a subject is lying, the load change information being acquired based on a detection signal corresponding to the load, and event information indicating an event causing a fluid balance of the subject associated with the change over time; By performing machine learning based on the learning data, a trained model that infers the event from the load change information is generated.

[0011] A sixth aspect for achieving the above object is a trained model generation method, inputting, as learning data, a combination of load change information indicating a change over time in a load applied to a bed on which a subject is lying, the load change information being acquired based on a detection signal corresponding to the load, and event information indicating an event causing a fluid balance of the subject, associated with the change over time; By performing machine learning based on the learning data, a trained model that infers the event from the load change information is generated.

[0012] According to the configurations relating to each of the fourth to sixth aspects described above, it is possible to provide a trained model that can not only reduce the burden associated with the prior design of explicit rules for the inference of causative events of fluid balance performed by the inference device relating to each of the first to third aspects, but also improve the generalization performance of the inference device. [Brief description of the drawings]

[0013] [Figure 1] 2 illustrates an example of a functional configuration of a moisture balance management system according to an embodiment. [Diagram 2] 2 illustrates an example of load change information acquired by the inference device of FIG. 1. [Diagram 3] 2 shows an example of information output by the output device of FIG. 1. [Figure 4] 2 illustrates an example of a functional configuration of a model generating device according to an embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Exemplary embodiments are described in detail below with reference to the accompanying drawings.

[0015] 1 illustrates a functional configuration of a fluid balance management system 10 (hereinafter, abbreviated as the management system 10) according to one embodiment. The management system 10 is configured to manage fluid balance information based on the fluid intake and excretion of a subject.

[0016] The management system 10 includes a detection device 11. The detection device 11 is configured to detect a load applied to a bed 20 on which a subject sits, and output a detection signal DT corresponding to the load. That is, the detection signal DT corresponds to the weight of the subject sitting on the bed 20.

[0017] The detection signal DT may be output continuously or intermittently at predetermined intervals. The output detection signal DT may be an analog signal or a digital signal.

[0018] The detection device 11 may take any suitable form. As an example, the detection device 11 may be configured to include load meters (so-called bed scales) arranged between each leg of the bed 20 and the floor. In this case, the detection device 11 generates the detection signal DT based on the load detected by each load meter. As another example, the detection device may be a load sensor built into the mattress or legs of the bed 20. In this case, the detection device 11 generates the detection signal DT based on the load detected by the load sensor.

[0019] The management system 10 includes an inference device 12. The inference device 12 is configured to obtain load change information indicating a change over time in the load applied to the bed 20 based on the detection signal DT, and infer an event that causes the subject's fluid balance from the load change information. Examples of the event include eating, drinking, excretion (including bedwetting), insensible perspiration, etc.

[0020] Specifically, the inference device 12 includes an input interface 121. The input interface 121 is configured to receive the detection signal DT output from the detection device 11. When the detection signal DT is an analog signal, the input interface 121 includes an appropriate conversion circuit including an A / D converter.

[0021] The inference device 12 includes a processor 122. The processor 122 is configured to acquire load change information LV, as exemplified in Fig. 2, based on the detection signal DT received by the input interface 121. The horizontal axis represents the passage of time. The vertical axis represents the load applied to the bed 20. Periods in which the load disappears, as seen in various places, represent situations in which the subject gets out of bed 20.

[0022] The processor 122 is configured to infer events that cause the subject's water balance based on the load change information LV. Some examples of inferences are given below.

[0023] During the period T, the load gradually decreases without a corresponding change in getting out of bed. If such a change is confirmed, the processor 122 infers insensible perspiration as the event causing the fluid balance.

[0024] At time t1, the load shows a significant increase after a change corresponding to getting out of bed. If such a change is identified, processor 122 infers a meal as the event causing the fluid balance.

[0025] At time t2, the load shows a significant decrease after showing a change corresponding to getting out of bed. If such a change is confirmed, the processor 122 infers excretion as an event causing fluid balance.

[0026] The inference device 12 includes an output interface 123. The processor 122 is configured to output inference information PR corresponding to the inference result from the output interface 123. The inference information PR is configured to include an inferred event and the time at which the event occurred. The inference information PR may be output in the form of analog data or in the form of digital data. When the inference information PR is output in the form of analog data, the output interface 123 includes an appropriate conversion circuit including a D / A converter.

[0027] The management system 10 includes an output device 13. The output device 13 may be a stationary device installed on the bed 20, or may be a mobile device that can be carried by the user. The inference information PR output from the output interface 123 is transmitted to the output device 13. The inference information PR may be transmitted via wired communication or wireless communication. The output device 13 is configured to output the inference result by the inference device 12 based on the inference information PR. The output by the output device 13 may take any suitable form involving at least one of presenting the inference result to the user and recording the inference information PR in a storage device (not shown).

[0028] According to this configuration, the weight change accompanying the fluid balance of the subject while getting out of bed 20 is reflected in the change in the load applied to the bed 20 when the subject returns to bed. Since the event causing the fluid balance is inferred based on the change, the management of the fluid balance while getting out of bed can be automated. This allows the fluid balance while getting out of bed, which was previously excluded from management, to be easily included in the management, improving the manageability of the fluid balance information of the subject not only while in bed but also while getting out of bed. In addition, the complementary recording of the fluid balance while getting out of bed, which was previously generally done manually, can be automated, reducing the burden on the user.

[0029] In order to improve the visibility of the managed moisture balance information, the output device 13 may be configured to output an image IM as exemplified in FIG. 3. The image IM displays an indicator corresponding to the inference result by the inference device 12 together with the load change information LV. In this example, the indicator includes a marker indicating the time when it is inferred that an event has occurred, and an icon indicating the inferred event. In this example, the icon includes a character indicating the event. In addition to or instead of the character, the event may be identified by a color, a symbol, a figure, or the like.

[0030] The output device 13 may output weight change information indicating a change in the subject's weight accompanying the occurrence of an event related to the inference, in addition to or instead of the information corresponding to the inference information PR. Specifically, the processor 122 of the inference device 12 specifies the load change information LV as the amount of change in the load applied to the bed 20 before and after the occurrence of the event included in the inference information PR. As described above, the load applied to the bed 20 corresponds to the subject's weight, so the amount of change in the load can be used as weight change information.

[0031] According to such a configuration, the recording of the subject's weight change due to fluid balance can be automated, further reducing the burden on the user in managing fluid balance information.

[0032] 1, the inference device 12 may include a trained model 124. The processor 122 may be configured to perform a process of inputting the load change information LV to the trained model 124 to obtain an inference result of an event that causes the subject's water balance.

[0033] The trained model 124 is an inference algorithm generated through machine learning using a neural network, which will be described later. The trained model 124 is configured to receive load change information LV, which indicates a change over time in the load applied to the bed 20, as an input, and to output an event that causes the subject's water balance as an inference result.

[0034] The trained model 124 is generated by a model generation device 30 illustrated in Fig. 4. That is, the model generation device 30 is configured to generate the trained model 124 to be installed in the inference device 12 described above.

[0035] The model generating device 30 includes an input interface 31. The input interface 31 is configured to receive learning data LG. The learning data LG includes load change information LV acquired for a subject and event information EV indicating an event that causes the water balance of the subject.

[0036] The model generation device 30 includes a processor 32. The processor 32 is configured to train a neural network using the training data LG, thereby generating a trained model 124. In this example, a method related to supervised learning is used as the process of training the neural network.

[0037] Specifically, the learning data LG is provided with teacher data that teaches what kind of load change information LV is acquired when a water balance occurs in the subject based on what kind of event.

[0038] The learning data LG may include information related to the weight change of the subject, such as temperature, humidity, and time.

[0039] The model generating device 30 includes an output interface 33. The output interface 33 is configured to output the trained model 124 generated by the processor 32 in a form that can be implemented in the inference device 12.

[0040] Such a configuration not only reduces the burden of prior design of explicit rules for the inference of causal events of fluid balance performed by the inference device 12, but also improves the generalization performance of the inference device 12.

[0041] The trained model 124 may be configured to be retrainable. For example, if the weight change of the subject during the period T in FIG. 2 is caused by bedwetting rather than insensible perspiration, the inference result displayed in the indicator illustrated in FIG. 3 is incorrect. In such a case, the inference information PR may be corrected as necessary.

[0042] As illustrated in Fig. 4, correction information CR corresponding to the correction can be input to the input interface 31 of the model generation device 30. In this example, the correction information CR is provided as a combination of the load change information LV in which the erroneous inference was made and event information corresponding to the corrected event "bedwetting." The processor 32 of the model generation device 30 executes re-learning of the trained model 124 so that the correction information CR is reflected. The timing of the re-learning can be set appropriately.

[0043] The re-learning of the trained model 124 may be configured to be executed by the processor 122 of the inference device 12. In this case, the input interface 121 of the inference device 12 is configured to receive the revision information CR.

[0044] With this configuration, the inference device 12 can improve the accuracy of inferring causative events of fluid balance based on information acquired from the subject after the management system 10 starts operating.

[0045] Each of the processor 122 of the inference device 12 and the processor 32 of the model generating device 30 having the various functions described above can be realized by a general-purpose microprocessor operating in cooperation with a general-purpose memory. Examples of the general-purpose microprocessor include a CPU, an MPU, and a GPU. Examples of the general-purpose memory include a ROM and a RAM. In this case, the ROM can store a computer program that executes the above-mentioned processing. The ROM is an example of a non-transitory computer-readable medium that stores a computer program. The general-purpose microprocessor specifies at least a part of the program stored on the ROM, expands it on the RAM, and executes the above-mentioned processing in cooperation with the RAM. The computer program may be pre-installed in the general-purpose memory, or may be downloaded from an external server via a communication network and then installed in the general-purpose memory. In this case, the external server is an example of a non-transitory computer-readable medium that stores a computer program.

[0046] Each of the processor 122 of the inference device 12 and the processor 32 of the model generation device 30 having the various functions described above may be realized by a dedicated integrated circuit capable of executing the above computer program, such as a microcontroller, an ASIC, or an FPGA. In this case, the above computer program is pre-installed in a storage element included in the dedicated integrated circuit. The storage element is an example of a computer-readable medium storing a computer program. Each of the processor 122 of the inference device 12 and the processor 32 of the model generation device 30 having the various functions described above may also be realized by a combination of a general-purpose microprocessor and a dedicated integrated circuit.

[0047] The above-described embodiment is merely an example for facilitating understanding of the present invention. The configurations according to the above-described embodiment may be appropriately modified or improved without departing from the spirit and scope of the present invention.

[0048] In the above embodiment, the model generating device 30 generates the trained model 124 through supervised learning. However, the model generating device 30 may generate the trained model 124 through unsupervised learning. In this case, the training data LG includes load change information indicating a time-dependent change in the load applied to the bed 20 while the subject is in bed. The trained model 124 generated through this training infers that an event related to water balance has occurred when the time-dependent change in the load indicated by the detection signal DT received from the detection device 11 deviates from the time-dependent change in the trained load.

[0049] In the above embodiment, the model generation device 30 generates the trained model 124 through machine learning using a neural network. However, the trained model 124 may be generated through other machine learning algorithms. Examples of other machine learning algorithms include decision trees, random forests, and support vector machines. [Explanation of symbols]

[0050] 10: Moisture balance management system, 11: Detection device, 12: Inference device, 121: Input interface, 122: Processor, 124: Trained model, 13: Output device, 20: Bed, 30: Model generation device, 31: Input interface, 32: Processor, DT: Detection signal, IM: Image, LG: Learning data, LV: Load change information, PR: Inference information

Claims

1. A detection device that outputs a detection signal corresponding to a load applied to a bed on which a subject is lying; an inference device that acquires load change information indicating a change over time of the load based on the detection signal, and infers an event that causes a fluid balance of the subject on the bed from the load change information; an output device that outputs inference information corresponding to an inference result of the event; Equipped with Moisture balance management system.

2. The inference device infers an event that causes fluid balance of the subject while the subject is out of bed based on the load change information when the subject returns to bed after getting out of bed. The fluid balance management system according to claim 1 .

3. the output device outputs an image displaying an index corresponding to the inference result together with the load change information. The fluid balance management system according to claim 1 or 2.

4. the output device outputs weight change information indicating a weight change amount of the subject together with the inference information. The fluid balance management system according to any one of claims 1 to 3.

5. The inference device infers the event using a trained model by machine learning. The fluid balance management system according to any one of claims 1 to 4.

6. The trained model is configured to be retrained so that a record of corrections to the inference result is reflected in the inference of the event. The fluid balance management system according to claim 5.

7. an interface that receives load change information indicating a change over time in a load applied to a bed on which a subject is lying, the change being acquired based on a detection signal corresponding to the load; a processor for inferring an event that causes fluid balance of the subject on the bed from the load change information and outputting information corresponding to the inference result of the event to an output device; Equipped with Reasoning device.

8. A computer program executable by a processor installed in an inference device, When executed, the inference device: receiving load change information indicating a change over time in a load applied to a bed on which a subject is lying, the change being acquired based on a detection signal corresponding to the load; Inferring from the load change information an event that causes fluid balance of the subject on the bed; outputting information corresponding to the inference result of the event to an output device; Computer program.

9. an interface that receives, as learning data, a combination of load change information indicating a change over time in a load applied to a bed on which a subject is lying, the load change information being acquired based on a detection signal corresponding to the load, and event information indicating an event that causes fluid balance of the subject on the bed, associated with the change over time; A processor that performs machine learning based on the learning data to generate a learned model that infers the event from the load change information; Equipped with Trained model generator.

10. The machine learning is performed by supervised learning. The trained model generating device according to claim 9 .

11. A computer program executable by a processor installed in a trained model generation device, By executing the above, the trained model generation device A combination of load change information indicating a change over time in a load applied to a bed on which a subject is lying, the change being acquired based on a detection signal corresponding to the load, and event information indicating an event causing a fluid balance of the subject on the bed, associated with the change over time, is received as learning data; generating a trained model for inferring the event from the load change information by performing machine learning based on the training data; Computer program.

12. inputting, as learning data, a combination of load change information indicating a change over time in a load applied to a bed on which a subject is lying, the load change information being acquired based on a detection signal corresponding to the load, and event information indicating an event causing fluid balance of the subject on the bed, associated with the change over time; generating a trained model for inferring the event from the load change information by performing machine learning based on the training data; How to generate trained models.

Images