Information processing method, information processing device, information processing program, method for generating machine learning completion model, and machine learning completion model

Abstract

Provided are an information processing method, an information processing device, an information processing program, a method for generating a machine learning completion model, and a machine learning completion model, which are capable of facilitating management of a subject of biological information. A processor (11) according to the present invention performs: a process for acquiring measurement-related information relating to the actual measurement performance of biological information for a prescribed amount of time by a biological information measurement device for a subject to be measured; deriving measurement tendency information on the basis of the measurement-related information, the measurement tendency information being information indicating the magnitude of the likelihood that the subject will continue to measure biological information during a future period after the prescribed period; and performing processing based on the measurement tendency information.

Description

Technical Field

[0001] The present invention relates to an information processing method, an information processing device, an information processing program, a method for generating a machine learning completion model, and a machine learning completion model. Background Art

[0002] For health management, it is sometimes necessary to continuously measure biological information such as weight, blood pressure, and blood sugar. However, depending on the person being measured, forgetting to measure this information can be a problem. Patent Document 1 describes a technology that acquires data from a user's biological information and, based on this data, moves the user's chess piece forward in Sugoroku, thereby increasing the user's motivation to measure their biological information.

[0003] Prior art literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent Application Publication No. 2019-3569 Summary of the Invention

[0006] Problems to be solved by the invention

[0007] Medical professionals such as doctors and nurses need to manage their patients and ensure they continue to measure their biological information. However, it is difficult for medical professionals who do not work with their patients to determine whether their patients are continuing to measure their biological information.

[0008] The purpose of the present disclosure is to provide an information processing method, an information processing device, an information processing program, a method for generating a machine learning completion model, and a machine learning completion model that can assist in the management of a subject's biological information.

[0009] Solutions for solving problems

[0010] The technology disclosed herein is as follows: In addition, although corresponding components and the like in the following embodiments are shown in parentheses, the present invention is not limited thereto. (1)

[0012] An information processing method, wherein a processor (processor 11) performs the following processing:

[0013] Acquiring measurement-related information related to actual results of measurements of biological information of a subject performed by a biological information measurement device for a predetermined period of time;

[0014] deriving measurement tendency information based on the measurement-related information, the measurement tendency information being information indicating a degree of probability that the subject will continue to measure biological information in a future period after the predetermined period;

[0015] Processing based on the above-mentioned measurement tendency information is performed.

[0016] According to (1), if the measurement-related information of the measured person can be obtained, it is possible to determine the degree of probability that the measured person will continue to measure his or her biometric information in the future. As a result, if it is determined that the probability of continuing to measure the biometric information is low, measures such as urging the measured person to measure his or her biometric information can be taken, thereby increasing the probability that the measured person will continue to measure his or her biometric information. (2)

[0018] The information processing method according to (1), wherein:

[0019] The processor inputs the measurement-related information into a model completed by machine learning, and acquires the measurement tendency information from the model.

[0020] According to (2), by periodically training the machine learning model, the accuracy of deriving the measurement tendency information can be improved, thereby enabling more appropriate management of the subjects. (3)

[0022] The information processing method according to (2), wherein:

[0023] The measurement results include the measurement values of the biological information.

[0024] According to (3), the measurement tendency information of the subject can be derived by collecting the measurement values measured by the subject, and thus the measurement tendency information can be easily derived without performing any special work on the subject. (4)

[0026] The information processing method according to (3), wherein:

[0027] The measurement-related information includes a representative value of the measurement values within the predetermined period.

[0028] According to (4), the measurement tendency information of the subject can be derived by collecting the measurement values measured by the subject, and thus the measurement tendency information can be easily derived without performing any special work on the subject. (5)

[0030] The information processing method according to (3), wherein:

[0031] The measurement-related information includes information indicating a change tendency of the measurement value within the predetermined period.

[0032] According to (5), the measurement tendency information of the subject can be derived by collecting the measurement values measured by the subject, so the measurement tendency information can be easily derived without performing any special work on the subject. (6)

[0034] The information processing method according to (2), wherein:

[0035] The above measurement results include the measurement time points of the biological information within the above specified period,

[0036] The measurement-related information includes information indicating characteristics of the distribution of the measurement time points.

[0037] According to (6), the measurement tendency information of the subject can be derived by collecting the measurement time points at which the biological information is measured from the subject. Therefore, the measurement tendency information can be easily derived without performing special operations on the subject. (7)

[0039] The information processing method according to (6), wherein

[0040] The information includes the elapsed time from the time when the biological information was last measured within the predetermined period to the reference time point.

[0041] According to (7), the measurement tendency information of the subject can be derived by collecting the measurement time points at which the biological information is measured from the subject. Therefore, the measurement tendency information can be easily derived without performing special operations on the subject. (8)

[0043] The information processing method according to (6), wherein

[0044] The information includes the number of times the biological information was measured during a period from the reference time point to a predetermined time before, which is a part of the predetermined period.

[0045] According to (8), the measurement tendency information of the subject can be derived by collecting the measurement time points at which the biological information is measured from the subject. Therefore, the measurement tendency information can be easily derived without performing special operations on the subject. (9)

[0047] The information processing method according to any one of (2) to (8), wherein

[0048] The above processor performs the following processing:

[0049] Acquiring information about the living conditions of the above-mentioned person being measured;

[0050] Furthermore, the measurer information is input into the model, and the measurement tendency information is obtained from the model.

[0051] According to (9), the measurement tendency information of the measurement subject can be derived using the measurer information and the measurement-related information of the measurement subject, and thus the accuracy of deriving the measurement tendency information can be improved. (10)

[0053] The information processing method according to (9), wherein:

[0054] The measurement subject information includes the time when the measurement subject last visited a hospital within the prescribed period, whether the measurement subject has any cohabiting persons, or the name of the disease the measurement subject suffers from.

[0055] According to (10), the measurement tendency information can be derived based on the living conditions of the person being measured, and thus the accuracy of deriving the measurement tendency information can be improved. (11)

[0057] The information processing method according to any one of (2) to (8), wherein

[0058] The above processor performs the following processing:

[0059] obtaining the model information of the biological information measuring device used by the subject;

[0060] Furthermore, the device model information is input into the model, and the measurement tendency information is obtained from the model.

[0061] According to (11), the measurement tendency information can be derived according to the model of the biological information measurement device used by the subject, and thus the accuracy of deriving the measurement tendency information can be improved. (12)

[0063] The information processing method according to any one of (2) to (11), wherein:

[0064] The model is generated by learning using measurement-related information and intervention information as learning data, wherein the measurement-related information is information related to the performance of measurements of the subject's biological information for a predetermined period of time, and the intervention information is information indicating whether an intervention to prompt the subject to undergo measurement of the biological information has been performed after the predetermined period.

[0065] The processor performs the following processing:

[0066] inputting the acquired measurement-related information and intervention information indicating that an intervention has been performed into the model, and performing the processing based on the measurement tendency information acquired from the model;

[0067] The acquired measurement-related information and no-intervention information indicating that no intervention was performed are input into the model, and the processing based on the measurement tendency information acquired from the model is performed.

[0068] According to (12), for example, the measurement tendency information in the case of intervention and the measurement tendency information in the case of no intervention can be compared, and this comparison can assist in determining whether intervention should be performed on the subject. (13)

[0070] The information processing method according to (12), wherein:

[0071] The intervention information included in the learning data indicating that the intervention was performed includes information on a time period during which the intervention was performed.

[0072] The processor performs a process of inputting the acquired measurement-related information and information on intervention in a specific time period into the model a plurality of times while changing the time period, and performs the process based on the measurement tendency information output from the model through the plurality of processes.

[0073] According to (13), for example, the measurement tendency information for each time period of intervention can be compared, and through this comparison, it can be determined in which time period the intervention is effective when performing intervention on the subject. (14)

[0075] The information processing method according to (12), wherein:

[0076] The intervention information included in the learning data indicating that the intervention was performed includes information on the content of the performed intervention.

[0077] The processor performs a process of inputting the acquired measurement-related information and the information of the intervention content into the model a plurality of times while changing the information of the content, and performs the process based on the measurement tendency information output from the model through the plurality of processes.

[0078] According to (14), for example, the measurement tendency information can be compared for each intervention content, and through this comparison, it can be determined what content of intervention is effective when intervening with the subject. (15)

[0080] An information processing device (information processing server 10) includes a processor (processor 11) that performs the following processing:

[0081] Acquiring measurement-related information related to actual results of measurements of biological information of a subject performed by a biological information measurement device for a predetermined period of time;

[0082] deriving measurement tendency information based on the measurement-related information, the measurement tendency information being information indicating a degree of probability that the subject will continue to measure biological information in a future period after the predetermined period;

[0083] Processing based on the above-mentioned measurement tendency information is performed. (16)

[0085] An information processing program causes a processor (processor 11) to execute the following steps:

[0086] Acquiring measurement-related information related to actual results of measurements of biological information of a subject performed by a biological information measurement device for a predetermined period of time;

[0087] deriving measurement tendency information based on the measurement-related information, the measurement tendency information being information indicating a degree of probability that the subject will continue to measure biological information in a future period after the predetermined period; and

[0088] Processing based on the above-mentioned measurement tendency information is performed. (17)

[0090] A method for generating a machine learning completion model, wherein the method for generating the machine learning completion model comprises a processor (processor 11) performing the following processing:

[0091] Acquiring a plurality of the following two types of information as learning data: one type of information is measurement-related information related to actual performance of biological information measurements of a subject by a biological information measurement device during a predetermined period (period T1) within a predetermined period of time in the past; and the other type of information is information regarding whether the biological information of the subject has been continuously measured during a period (period T2) after the predetermined period within the predetermined period;

[0092] The program is caused to execute machine learning based on the above-mentioned multiple learning data to generate a machine learning completion model (machine learning completion model 13). The above-mentioned machine learning completion model (machine learning completion model 13) is a model that outputs measurement tendency information when measurement-related information related to the measurement performance of the biological information of the subject performed by the biological information measurement device for a specified period of time is input. The above-mentioned measurement tendency information is information indicating the possibility that the subject will continue to measure the biological information in the future period after the specified period. (18)

[0094] A machine learning completion model (machine learning completion model 13) is obtained by performing machine learning using the following two types of information as learning data: one type of information is measurement-related information related to actual performance of biological information measurements of a subject by a biological information measurement device during a predetermined period (period T1) within a predetermined period of time in the past; and the other type of information is information regarding whether the biological information of the subject has been continuously measured during a period (period T2) after the predetermined period within the predetermined period.

[0095] The above-mentioned machine learning completion model (machine learning completion model 13) causes the processor (processor 11) to perform the following processing:

[0096] Measurement-related information related to the measurement performance of the biological information of the person being measured for a specified period of time is used as input to output measurement tendency information, which is information indicating the likelihood that the person being measured will continue to measure the biological information in the future period after the specified period.

[0097] Effects of the Invention

[0098] According to the present disclosure, it is possible to facilitate management of the biological information of the person being measured. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] Figure 1 It is a schematic diagram showing the schematic configuration of the management system 100 .

[0100] Figure 2 It is a figure which shows typically the measurement data.

[0101] Figure 3 This is a flowchart used to explain the method of generating a machine learning completion model.

[0102] Figure 4 FIG. 1 is a diagram schematically showing measurement data of user X. FIG.

[0103] Figure 5 It is a diagram showing a modified example of the management system 100 .

[0104] Figure 6 It is a diagram showing an example of a screen displayed on the display device of the facility terminal 40.

[0105] Figure 7 It is a diagram showing another example of a screen displayed on the display device of the facility terminal 40 . DETAILED DESCRIPTION

[0106] (Overview of the Information Processing Method of the Present Disclosure)

[0107] An information processing method is a method in which a processor performs the following processing: obtaining measurement-related information related to the actual measurement results of the biological information of a person being measured, performed by a biological information measuring device such as a weight scale, a blood pressure monitor, or a blood glucose meter, for a specified period of time; deriving measurement tendency information based on the measurement-related information, the measurement tendency information being information indicating the likelihood that the person being measured will continue to measure the biological information in the future after the specified period; and performing processing based on the measurement tendency information.

[0108] The measurement-related information preferably includes information indicating the magnitude or trend of changes in the biometric information measurement values within the aforementioned predetermined period, or information indicating characteristics of the distribution of biometric information measurement time points within the aforementioned predetermined period. The measurement trend information preferably includes information indicating the probability that the subject will continue to have their biometric information measured in the future. In this way, based on the subject's actual biometric information measurement performance, it is possible to predict the degree to which the subject will continue to measure their biometric information in the future, thereby enabling appropriate recommendations to be made to the subject, thereby encouraging the subject to continue measuring their biometric information.

[0109] Hereinafter, a configuration example of a management system including a device that executes the information processing method of the present disclosure will be described.

[0110] (System Configuration)

[0111] Figure 1 This is a schematic diagram showing the general configuration of a management system 100. Management system 100 is a system for assisting individuals (hereinafter referred to as users) in continuously measuring biological information such as weight, blood pressure, pulse, and blood sugar. Management system 100 includes an information processing server 10, a measurement data management server 20, a facility terminal 40, and multiple user terminals 50, all of which are connected to a network 30 such as the Internet.

[0112] The user terminal 50 is an electronic device such as a smartphone held by the user. A biological information measuring device, such as a weight scale, blood pressure monitor, pulse rate monitor, or blood glucose meter, held by the user, is communicatively connected to the user terminal 50. Measurement data obtained by the biological information measuring device is transmitted from the user terminal 50 to the measurement data management server 20. The measurement data includes biological information measurement values such as weight, blood pressure, pulse rate, or blood glucose level, as well as information on the measurement date and time. The following describes an example in which the biological information measuring device is a blood pressure monitor and the measured value is a blood pressure value (preferably the highest blood pressure).

[0113] The measurement data management server 20 associates the measurement data transmitted from the user terminal 50 with information identifying the user and stores the information in a database, thereby managing the measurement data for each user. Figure 2It is a figure which shows typically the measurement data. Figure 2 Graph 2 shows the measurement data D1 of user A.

[0114] The measurement data includes a plurality of sets of measurement dates and times and measurement values (blood pressure values) measured at the measurement dates and times. Figure 2 In the example, the maximum blood pressure value, which is the measured value, is displayed in a bar graph. Dates and times when no bar graph is present indicate dates and times when no measurement was performed. This measurement data is collected from multiple users over a long period of time and stored in a database.

[0115] The measurement data management server 20 includes a sample data set, which serves as the target for extracting teaching data used to generate the machine learning model 13, described later. Each measurement data set included in this sample data set includes, for example, measurement results (including the times and values of measurements) within a predetermined period T1, beginning with the date the user registered their use of the measurement data management server 20, and measurement results within a predetermined period T2, beginning with the day following the end of period T1. While periods T1 and T2 are not particularly limited, period T2 is set to be longer than period T1. For example, period T1 is 30 days, and period T2 is 90 days.

[0116] The facility terminal 40 is an electronic device such as a personal computer, smartphone, or tablet terminal installed in a medical facility such as a hospital. By accessing the measurement data management server 20 from the facility terminal 40, measurement data for a specific user can be downloaded to the facility terminal 40 for reference. The facility terminal 40 includes, for example, a display device such as an organic EL (electroluminescence) display or a liquid crystal display, as well as a speaker.

[0117] The information processing server 10 includes a processor 11 and a storage unit 12. The storage unit 12 includes working memory such as RAM (Random Access Memory) and non-transitory storage media such as a hard disk or flash memory. The storage unit 12 stores an information processing program for the information processing server 10 to execute the information processing method.

[0118] The processor 11 is a general-purpose processor such as a CPU (Central Processing Unit) that executes software (programs) to perform various functions; a programmable logic device (PLD) such as an FPGA (Field Programmable Gate Array) that is a processor whose circuit configuration can be modified after manufacturing; or a dedicated circuit such as an ASIC (Application Specific Integrated Circuit) that is a processor with a circuit configuration specifically designed to perform specific processing. The processor 11 can be composed of a single processor or a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a CPU and an FPGA). More specifically, the hardware structure of the processor 11 is a circuit composed of a combination of circuit elements such as semiconductor devices. If the processor 11 is composed of multiple processors, the multiple processors do not need to be located in the same device and can be located in multiple devices distributed via the network 30.

[0119] The storage unit 12 stores a machine learning model 13. The machine learning model 13 is generated by causing a learning model composed of a program to perform machine learning using teaching data. The machine learning model 13 can be generated by the processor 11 of the information processing server 10 or by a processor of a computer different from the information processing server 10. Below, the method for generating the machine learning model 13 is described, assuming that the processor 11 generates the machine learning model 13.

[0120] (Method for generating a machine learning model)

[0121] Figure 3 This is a flowchart used to explain the method of generating a machine learning completion model.

[0122] The processor 11 acquires the sample data set described above from the measurement data management server 20 (step S1). Next, the processor 11 derives measurement-related information (in other words, information indicating characteristics of the measurement results) related to the blood pressure measurement results during period T1 based on the data for period T1 within each measurement data set in the acquired sample data set (step S2).

[0123] The measurement-related information includes, for example, first information indicating characteristics of the measurement values in the period T1 or second information indicating characteristics of the distribution of measurement time points in the period T1.

[0124] The first information is, for example, a representative value of the blood pressure values during period T1. The representative value may be, for example, the average value of the blood pressure values measured during period T1, the average value of the blood pressure values measured during period T1 excluding the minimum and maximum values, or the median value of the blood pressure values measured during period T1.

[0125] As another example, the first information is information indicating the variation trend of the blood pressure value in the period T1. The information indicating the variation trend is, for example, information derived by the least square method or the like. Figure 2 The straight line L1 showing the time series change of the blood pressure value in the period T1 of the measurement data D1 is the slope of the straight line L1. The first information may also be a value indicating the change of the blood pressure value in the period T1. Figure 2 Image of the chart shown.

[0126] The second information, for example, indicates whether the number of blood pressure measurements has decreased during a period close to period T2 within period T1. For example, the second information may be divided into multiple groups, and the number of blood pressure measurements in the group closest to period T2 (the period from the end of period T1 to a predetermined time) may be used. As another example, the second information may be the time elapsed from the last blood pressure measurement within period T1 to the end of period T1.

[0127] Next, the processor 11 derives measurement continuation information indicating whether the user has continuously measured blood pressure values during period T2 based on the data for period T2 in each measurement data set acquired from the sample data set (step S3). The user's continuous blood pressure measurement indicates that the number of blood pressure measurements during period T2 is a predetermined value (three, for example) or greater. If the number of blood pressure measurements during period T2 is the predetermined value or greater, the measurement continuation information indicates that the measurement has been continued (e.g., "1," "True"). If the number of blood pressure measurements during period T2 is less than the predetermined value, the measurement continuation information indicates that the measurement has not been continued (e.g., "0," "False").

[0128] The processor 11 uses the measurement-related information and measurement continuation information derived based on each measurement data as a data set of teaching data, and causes the learning model to execute machine learning based on these multiple data sets, thereby generating a machine learning completed model 13 .

[0129] The machine learning completion model 13 is a model that, when input with measurement-related information related to actual blood pressure measurement results of a user using a blood pressure monitor over a predetermined period (a period of the same length as period T1), learns various parameters to output measurement tendency information indicating the likelihood that the user will continue to measure blood pressure values in a future period (a period of the same length as period T2) following the predetermined period. The measurement tendency information is preferably information indicating the probability that the user will continue to measure blood pressure values (or the probability that the user will not continue to measure blood pressure values) in the future period.

[0130] Thus, the machine learning model 13 is a model obtained by machine learning in the following manner: based on a large amount of measurement-related information and the corresponding measurement continuation information, the model estimates and outputs the probability that a specific user will continue to perform measurements in the future based on the measurement-related information obtained from the user's past measurement data. The machine learning method is not particularly limited; for example, any method such as logistic regression, decision tree, random forest, gradient boosting decision tree, neural network, etc. can be used.

[0131] Statistical analysis of a vast amount of previously acquired measurement data reveals that users whose blood pressure tends to be high during period T1 (users with high representative blood pressure values) tend to discontinue measuring their blood pressure during the subsequent period T2. This is likely due to the fact that consistently measured blood pressure values are high, which reduces their motivation to continue measuring. Furthermore, the above results show that users whose blood pressure tends to rise during period T1 tend to discontinue measuring their blood pressure during the subsequent period T2. This is likely due to the fact that consistently measured blood pressure values tend to deteriorate rather than improve, which reduces their motivation to continue measuring.

[0132] Furthermore, according to the above results, users who took fewer measurements during the period close to period T2 within period T1 (e.g., the latter half of period T1 when divided into the first and second halves) tend not to continue measuring their blood pressure during the subsequent period T2. Furthermore, according to the above results, users whose last measurement in period T1 lasted a long time until the end of period T1 (users who had not taken a measurement in the recent period) tend not to continue measuring their blood pressure during the subsequent period T2. It should be noted that these considerations also apply to other biological information such as weight and blood sugar levels.

[0133] Therefore, by performing machine learning on a set of measurement-related information and measurement continuation information, a model can be generated that estimates the likelihood that a specific user will continue to measure blood pressure values in the future based on the measurement-related information obtained based on the measurement data of the specific user.

[0134] (Machine learning completes the use of models)

[0135] The processor 11 uses the machine learning completion model 13 generated in this way to estimate the degree of possibility that a user who can only obtain measurement data for period T3, which is the same length as period T1, will continue to measure blood pressure values in the future period after period T3 (a period of the same length as period T2), and performs processing based on the estimation result.

[0136] For example, Figure 4 As shown, assume the following situation: a healthcare professional wishes to know whether a specific user X, for whom measurement data DX has been obtained for a period T3, will continue to undergo measurements in the future period after period T3. The healthcare professional operates the facility terminal 40 to read user X's measurement data DX at the end of period T3, and transmits this measurement data DX to the information processing server 10, requesting the server to estimate the probability of user X continuing measurements in the future period after the end of period T3.

[0137] Upon receiving the measurement data DX, the processor 11 of the information processing server 10 derives measurement-related information related to the measurement results during period T3 based on the measurement data DX. For example, the processor 11 derives a representative value (such as an average or median value) of the blood pressure values during period T3 from the measurement data DX as the measurement-related information. It should be noted that the processor of the facility terminal 40 may also derive the measurement-related information based on the measurement data DX. In this case, the processor of the facility terminal 40 transmits the derived measurement-related information to the information processing server 10, requesting it to estimate the measurement continuation probability.

[0138] Next, the processor 11 inputs the measurement-related information derived by the processor 11 itself or the measurement-related information received from the facility terminal 40 into the machine learning completion model 13, and obtains measurement tendency information (preferably probabilistic information) indicating the likelihood that the user X will continue to measure blood pressure values in the future period after period T3 from the machine learning completion model 13.

[0139] Upon acquiring the measurement trend information, the processor 11 causes the display device of the facility terminal 40 to display the measurement trend information or outputs the measurement trend information as a voice from the speaker of the facility terminal 40. The processor 11 may also send the measurement trend information to the facility terminal 40 in a form such as an email.

[0140] For example, a healthcare provider can determine whether user X is likely to continue measuring blood pressure in the future by viewing the measurement trend information displayed on the display device of the facility terminal 40. For example, if the measurement trend information (probability of continued measurement) is low, the healthcare provider can proactively intervene with user X to prompt him to measure his blood pressure (using, for example, a notification via an application installed on user terminal 50, an email, or a phone call). This increases the likelihood that user X will continue measuring their blood pressure. By appropriately executing such interventions, it is expected that user X will improve his or her own health awareness, improve diagnostic accuracy by enabling detailed assessment of changes in user X's blood pressure, and contribute to improving user X's health.

[0141] It should be noted that each measurement data of the sample data group used to generate the machine learning completion model 13 is set to include the measurement date and time and the measurement value, but when the second information (information representing the characteristics of the distribution of the measurement time point) is used as the measurement-related information input to the machine learning completion model 13, each measurement data only needs to include at least the measurement date and time, and may not include the measurement value.

[0142] (Variation of the machine learning model)

[0143] The machine learning completion model 13 is a model obtained by machine learning using measurement-related information and measurement duration information as teaching data. However, machine learning can also be performed by further including information about the measurer in this teaching data. This measurer information refers to information related to the living conditions of the user who obtained each measurement data in the sample data set. Examples of this information include the time of the user's last hospital visit during period T1, information about whether the user has cohabiting relatives, or information about the name of the disease the user suffers from. In this case, the measurer information related to the living conditions of the user who obtained each measurement data is added to each measurement data in the sample data set stored in the database by the measurement data management server 20.

[0144] The processor 11 then uses as teaching data a combination of the measurer information included in each measurement data set in the sample data set and the measurement-related information obtained from each measurement data set, and uses as teaching data the measurement continuation information obtained from each measurement data set, and causes the learning program to perform machine learning based on this set of teaching data, thereby generating a machine learning completion model 13. In this manner, the machine learning completion model 13 is generated such that, when a set of measurer information and measurement-related information is input, the model outputs measurement tendency information corresponding to the set, indicating the likelihood that the user will continue to perform measurements in the future.

[0145] By further leveraging the user's measurement provider information, it's possible to more accurately estimate whether the user will continue measuring their biometric information in the future. For example, compared to users who didn't visit a hospital during period T1, the former has a higher health awareness and is therefore more likely to continue measuring. Furthermore, compared to users who live with others, the former, who tend to point out, for example, that they forgot to measure, are more likely to continue measuring. By learning these trends in continued measurement based on living conditions, it's possible to accurately estimate the likelihood of a user continuing measuring.

[0146] It should be noted that the aforementioned measurement user information can also be replaced with information about the model of the biometric measurement device used by the user for measurement. The tendency to continue measurement may also vary depending on the usability of the biometric measurement device (ease of communication, ease of measurement, etc.). Therefore, by learning this tendency based on the device model used, the likelihood of the user continuing measurement can be estimated with higher accuracy.

[0147] (First Modification of the Management System)

[0148] Figure 5 1 is a diagram showing a modified example of the management system 100. Figure 5 In the management system 100 shown, Figure 1 The difference is that the machine learning completed model stored in the storage unit 12 is changed into two machine learning completed models 13A and 13B.

[0149] exist Figure 5 In the illustrated management system 100, the sample data sets stored in the database by the measurement data management server 20 include an intervention sample data set and a non-intervention sample data set. The intervention sample data set consists of measurement data from users who underwent an intervention requiring blood pressure measurement during period T2, while the non-intervention sample data set consists of measurement data from users who did not undergo an intervention requiring blood pressure measurement during period T2. Each measurement data point in the intervention sample data set is associated with information indicating the presence of an intervention (e.g., "1"). Each measurement data point in the non-intervention sample data set is associated with information indicating the absence of an intervention (e.g., "0").

[0150] The machine learning model 13A is a model generated by performing machine learning on a teaching data set of measurement-related information and measurement continuity information generated based on each measurement data of the intervention sample data group, and constitutes a first model.

[0151] The machine-learned model 13B is a model generated by performing machine learning on a teaching data set of measurement-related information and measurement continuity information generated based on each measurement data of the non-intervention sample data set, and constitutes a second model.

[0152] exist Figure 5 In the management system 100 shown, the processor 11 will, for example, Figure 4 The measurement-related information derived from the measurement data DX of the user X is input to the machine learning completion model 13A and the machine learning completion model 13B, and processing is performed based on the measurement tendency information output from the machine learning completion model 13A and the machine learning completion model 13B.

[0153] For example, the processor 11 causes the display device of the facility terminal 40 to display both the measurement trend information outputted by the machine learning completion model 13A and the machine learning completion model 13B, or to display the difference between these two pieces of measurement trend information (probabilities) as the effect of the intervention. Specifically, the display device may display a message such as "The probability of continued measurement for user X in the future is 30%, but intervention can increase this probability to 60%." By clearly indicating the effects of the intervention, medical practitioners will not hesitate in deciding whether to intervene for user X.

[0154] It should be noted that a plurality of models for different time periods for intervention may be prepared and stored in advance in the storage unit 12 as the machine-learned completed model 13A. For example, a morning intervention model and an afternoon intervention model may be pre-generated as the machine-learned completed model 13A. The morning intervention model is a model generated by machine learning on a teaching data set of measurement-related information and measurement continuity information generated based on each measurement data set of a sample data set of users who underwent intervention in the morning (e.g., 9:00 to 12:00). The afternoon intervention model is a model generated by machine learning on a teaching data set of measurement-related information and measurement continuity information generated based on each measurement data set of a sample data set of users who underwent intervention in the afternoon (e.g., 12:00 to 17:00).

[0155] Then, the processor 11 will, for example, Figure 4 The measurement-related information derived from the measurement data DX of user X shown in FIG. 1 is input into the morning intervention model in the machine learning model 13A, the afternoon intervention model in the machine learning model 13A, and the machine learning model 13B, respectively, and processing based on the measurement tendency information output from each model is performed (for example, displaying Figure 6By this process, not only can the change in the probability of continuous measurement due to the presence or absence of intervention be known, but also whether the probability is higher in the morning or in the afternoon when intervention is performed can be known. For example, by Figure 6 In the example, in the morning, intervening in the user can further increase the possibility that the user will continue to measure blood pressure.

[0156] Furthermore, a plurality of models with different intervention contents may be prepared and stored in advance in the storage unit 12 as the machine-learned completed model 13A. For example, a telephone intervention model generated by machine learning on a teaching dataset of measurement-related information and measurement continuity information generated based on each measurement data set of a sample data set of users who underwent intervention using a telephone, and an application intervention model generated by machine learning on a teaching dataset of measurement-related information and measurement continuity information generated based on each measurement data set of a sample data set of users who underwent intervention using notifications from an application, may be generated as the machine-learned completed model 13A.

[0157] Then, the processor 11 will, for example, Figure 4 The measurement-related information derived from the measurement data DX of user X shown in FIG. 1 is input to the telephone intervention model in the machine learning completion model 13A, the application intervention model in the machine learning completion model 13A, and the machine learning completion model 13B, respectively, and processing based on the measurement tendency information output from each model is performed (for example, displaying Figure 7 By this process, not only the change of the continuous probability caused by the presence or absence of intervention can be known, but also how the probability changes according to the content of the intervention when intervention is performed. For example, by using the content with the best effect (in Figure 7 Intervention with the user (in this example, a phone call) can further increase the likelihood that the user will continue to measure blood pressure. It should be noted that intervention content may also include a message conveyed to the user during the intervention.

[0158] (Second Modification of the Management System)

[0159] The machine learning completion model 13A and the machine learning completion model 13B can also be implemented by a single machine learning completion model (denoted as 13C). Specifically, the machine learning completion model 13C can be a model generated by performing machine learning on a teaching data set consisting of measurement-related information generated based on each measurement data set in the sample data set, intervention information indicating the presence or absence of intervention associated with the measurement data, and measurement continuity information generated based on the measurement data.

[0160] The processor 11 is based on, for example Figure 4 The measurement data DX of the user X shown is used to derive measurement-related information, the measurement-related information and the information indicating intervention are input into the machine learning completion model 13C, the first measurement tendency information (the measurement continuation tendency in the case of future intervention) output from the machine learning completion model 13C is obtained, the measurement-related information and the information indicating no intervention are input into the machine learning completion model 13C, the second measurement tendency information (the measurement continuation tendency in the case of future non-intervention) output from the machine learning completion model 13C is obtained, and processing is performed based on the first measurement tendency information and the second measurement tendency information.

[0161] For example, the processor 11 causes the display device of the facility terminal 40 to display the first measurement trend information and the second measurement trend information together (see Figure 6 、 Figure 7 ), or the display device of the facility terminal 40 displays the difference between the first measurement trend information and the second measurement trend information (probability) as the effect of the intervention.

[0162] In the intervention information in the teaching data set used in the second modification, information on a time zone in which the intervention was performed (hereinafter referred to as intervention time zone information) may be added to the information on the presence of intervention.

[0163] In this case, the processor 11 is based on, for example Figure 4 The measurement data DX of the user X shown is used to derive measurement-related information, the measurement-related information and the information indicating that there is intervention in the morning are input into the machine learning completion model 13C, the third measurement tendency information (the measurement tendency in the case where intervention is performed in the morning in the future) output from the machine learning completion model 13C is obtained, the measurement-related information and the information indicating that there is intervention in the afternoon are input into the machine learning completion model 13C, the fourth measurement tendency information (the measurement tendency in the case where intervention is performed in the afternoon in the future) output from the machine learning completion model 13C is obtained, and processing is performed based on the third measurement tendency information and the fourth measurement tendency information.

[0164] For example, the processor 11 causes the display device of the facility terminal 40 to display the third measurement trend information and the fourth measurement trend information (see Figure 6 ), or the display device of the facility terminal 40 displays the difference between the third measurement trend information and the fourth measurement trend information (probability) as the effect of the different intervention time periods.

[0165] In addition, the information of the content of the intervention performed (intervention content information) may be used instead of the above-mentioned intervention time period information. In this case, the processor 11 may perform the following processing multiple times by changing the intervention content, for example: Figure 4The processor 11 derives measurement-related information from the measurement data DX of user X shown in FIG. This measurement-related information and information indicating specific intervention contents are input into the machine learning completion model 13C. The processor 11 acquires the plurality of pieces of measurement trend information (measurement trends for each intervention content when an intervention is performed in the future) output from the machine learning completion model 13C through these multiple processing steps and performs processing based on this plurality of pieces of measurement trend information.

[0166] For example, the processor 11 causes the display device of the facility terminal 40 to display a plurality of pieces of measurement tendency information at once (see Figure 7 ), or the display device of the facility terminal 40 displays the difference between these multiple measurement tendency information (probabilities) as the effect of different intervention contents.

[0167] In the description so far, the processor 11 has been configured to use machine learning to complete a model and derive measurement trend information from the measurement-related information. However, a table storing the correspondence between measurement-related information and measurement trend information may be pre-generated and stored in the storage unit 12. The processor 11 may then derive future measurement trend information for user X based on the measurement-related information derived based on user X's measurement data and this table.

[0168] For example, the measurement data of the sample data group is divided into multiple groups based on the size of the representative blood pressure values within period T1. Based on the measurement continuity information corresponding to the measurement data belonging to each group, the continuity probability is calculated for each group as (the number of users who have continuously measured / the total number of users in the group). The continuity probability thus calculated is simply associated with the group to generate the above table. When the measurement-related information of user X is obtained, the processor 11 simply determines the group to which the measurement-related information belongs and obtains the continuity probability corresponding to the determined group from the above table. When the model is completed using machine learning as described above, machine learning is performed each time sample data is stored, thereby improving the accuracy of the estimated continuity probability derived by the model. Therefore, it is more preferable to complete the model using machine learning.

[0169] While various embodiments have been described above, the present invention is not limited to the aforementioned examples. A person skilled in the art will readily conceive of various variations or modifications within the scope of the claims, and it should be understood that these variations or modifications also fall within the technical scope of the present invention. Furthermore, the various components of the aforementioned embodiments may be arbitrarily combined without departing from the spirit of the invention.

[0170] It should be noted that the present application is based on a Japanese patent application (Japanese Patent Application No. 2023-037755) filed on March 10, 2023, the contents of which are incorporated herein by reference.

[0171] Description of Reference Numerals

[0172] D1: measurement data;

[0173] T1, T2, T3: period;

[0174] L1: straight line;

[0175] 10: Information processing server;

[0176] 11: Processor;

[0177] 12: Storage unit;

[0178] 13, 13A, 13B: Machine learning completes the model;

[0179] 20: measurement data management server;

[0180] 30: Network;

[0181] 40: Facility terminal;

[0182] 50: user terminal;

[0183] 100: Management system.

Claims

1. An information processing method, wherein a processor performs the following processing: Acquiring measurement-related information related to actual results of measurements of biological information of a subject performed by a biological information measurement device for a predetermined period of time; deriving measurement tendency information based on the measurement-related information, the measurement tendency information being information indicating a degree of probability that the subject will continue to measure biological information in a future period after the predetermined period; Processing based on the measurement tendency information is performed.

2. The information processing method according to claim 1, wherein: The processor inputs the measurement-related information into a model completed by machine learning, and acquires the measurement tendency information from the model.

3. The information processing method according to claim 2, wherein: The measurement results include the measurement values of the biological information.

4. The information processing method according to claim 2, wherein: The measurement results include the measurement time points of the biological information within the predetermined period. The measurement-related information includes information indicating characteristics of the distribution of the measurement time points.

5. The information processing method according to any one of claims 2 to 4, wherein: The processor performs the following processing: Acquiring information about the person being measured that is related to their living conditions; Furthermore, the measurer information is input into the model, and the measurement tendency information is obtained from the model.

6. The information processing method according to any one of claims 2 to 4, wherein: The processor performs the following processing: acquiring model information of the biological information measuring device used by the subject; Furthermore, the device model information is input into the model, and the measurement tendency information is obtained from the model.

7. The information processing method according to any one of claims 2 to 4, wherein: The model is generated by learning using measurement-related information and intervention information as learning data, wherein the measurement-related information is information related to the performance of measurements of the subject's biological information for a predetermined period of time, and the intervention information is information indicating whether an intervention to prompt the subject to undergo measurement of the biological information has been performed after the predetermined period. The processor performs the following processing: inputting the acquired measurement-related information and intervention information indicating that an intervention has been performed into the model, and performing the processing based on the measurement tendency information acquired from the model; The acquired measurement-related information and no-intervention information indicating that no intervention was performed are input into the model, and the processing based on the measurement tendency information acquired from the model is performed.

8. The information processing method according to claim 7, wherein: The intervention information included in the learning data indicating that the intervention was performed includes information on a time period during which the intervention was performed. The processor performs a process of inputting the acquired measurement-related information and information on intervention in a specific time period into the model a plurality of times while changing the time period, and performs the process based on the measurement tendency information output from the model through the plurality of processes.

9. The information processing method according to claim 7, wherein: The intervention information included in the learning data indicating that the intervention was performed includes information on the content of the performed intervention. The processor performs a process of inputting the acquired measurement-related information and the information of the intervention content into the model a plurality of times while changing the information of the content, and performs the process based on the measurement tendency information output from the model through the plurality of processes.

10. An information processing device comprising a processor, wherein the processor performs the following processing: Acquiring measurement-related information related to actual results of measurements of biological information of a subject performed by a biological information measurement device for a predetermined period of time; deriving measurement tendency information based on the measurement-related information, the measurement tendency information being information indicating a degree of probability that the subject will continue to measure biological information in a future period after the predetermined period; Processing based on the measurement tendency information is performed.

11. An information processing program, the information processing program causing a processor to execute the following steps: Acquiring measurement-related information related to actual results of measurements of biological information of a subject performed by a biological information measurement device for a predetermined period of time; deriving measurement tendency information based on the measurement-related information, the measurement tendency information being information indicating a degree of probability that the subject will continue to measure biological information in a future period after the predetermined period; as well as Processing based on the measurement tendency information is performed.

12. A method for generating a machine learning completion model, wherein the method for generating a machine learning completion model comprises a processor performing the following processing: Acquiring a plurality of the following two types of information as learning data: one type of information is measurement-related information related to actual performance of measurements of the subject's biological information by the biological information measuring device during a predetermined period of a certain past period; and the other type of information is information regarding whether the subject's biological information has been continuously measured during a period after the predetermined period of the certain period; The program is caused to execute machine learning based on the plurality of learning data to generate a machine learning completion model, wherein the machine learning completion model is a model that outputs measurement tendency information when measurement-related information related to the measurement performance of the biological information of the subject performed by the biological information measuring device for a specified period of time is input, and the measurement tendency information is information indicating the likelihood that the subject will continue to measure the biological information in a future period after the specified period.

13. A machine learning completion model, the machine learning completion model performing machine learning using the following two types of information as learning data: one type of information is measurement-related information related to the performance of biometric information measurements of a subject by a biometric information measurement device during a specified period of a certain past period of time; the other type of information is information regarding whether the subject has continuously had their biometric information measured during a period after the specified period of time within the certain period of time; The machine learning completion model causes the processor to perform the following processing: Measurement-related information related to the measurement performance of the biological information of the person being measured for a specified period of time is used as input to output measurement tendency information, which is information indicating the likelihood that the person being measured will continue to measure the biological information in the future period after the specified period.

Images