JP2005328924A - Blood glucose level prediction device, blood glucose level prediction model creation device, and program - Google Patents

Abstract

An object of the present invention is to improve the prediction accuracy of blood glucose level without requiring an advanced examination technique used in clinical examinations.
A data acquisition unit 15A of a blood sugar level prediction model creation device 1 acquires, as a model input value, a physical condition variable value indicating the amount of energy acquired or consumed by a user and his / her physical condition determined subjectively by the user. Then, the history data generation means 15C generates a plurality of history data from a set of the model input value and the model output value composed of the blood glucose level measured by the user, and the model creation means 15D generates any arbitrary data based on the history data. A prediction model 14C that predicts the blood glucose level of the user from the input conditions is created.
[Selection] Figure 1

Description

  The present invention relates to a data prediction technique, and more particularly to a blood sugar level prediction technique for predicting a change in blood sugar level from a meal amount or exercise amount.

  In recent years, the number of diabetic patients has been increasing worldwide. Diabetes is a disease in which the blood glucose level indicating the concentration of glucose in the blood is high, that is, a disease in which hyperglycemia continues chronically. When glucose, which is an energy source, is taken into cells by the human body, a hormone called insulin is required. This insulin is a hormone produced by β cells of the pancreas, and when the amount of secreted insulin decreases or the sensitivity of the cells to insulin decreases, glucose overflows into the blood, resulting in hyperglycemia.

Treatment methods for diabetes include exercise therapy, diet therapy, oral hypoglycemic agent method, and insulin dose method, all of which maintain good blood glucose control over a long period of time and prevent complications The purpose is that. For this reason, many patients measure blood glucose using a blood-collecting self-blood glucose meter at early morning fasting or every meal for blood glucose control, and determine their exercise amount, meal amount, and insulin dose. Yes.
However, it is not easy for patients to use SMBG (Self Monitoring of Blood Glucose) data for glycemic control. Diabetes specialists consider the relevance and regularity of their lives based on their experience. Is currently underway.

  In view of the possibility of supporting blood glucose control more effectively by using data such as blood glucose level accumulated over a long period of time, it is related to the prediction of blood glucose level using data mining technology and further to the life management of diabetic patients. Techniques have been proposed (see, for example, Patent Document 1).

In general, factors affecting the change in blood glucose level include blood glucose level obtained from blood tests and glycohemoglobin (HbA _1C ), as well as lifestyle information such as diet and exercise. Among these, the amount of meal, the amount of exercise, and the blood glucose level can be easily collected with a commercially available device, but the clinical examination at a medical institution is necessary for glycated hemoglobin.
Therefore, in consideration of the fact that the user can easily collect blood sugar levels, the blood sugar level, the amount of meal (calories consumed), and the amount of exercise (calories burned) are used as input information in the conventional blood sugar level prediction technology. Based on this, the blood sugar level after several hours is predicted.

The applicant has not yet found prior art documents related to the present invention by the time of filing other than the prior art documents specified by the prior art document information described in this specification.
JP 2003-173375 A

However, in such a conventional blood glucose level prediction technique, objective input information acquired at a frequency of about once to several times a day, for example, blood glucose level, meal amount (calorie intake), and exercise amount (consumption) Since the increase / decrease in blood glucose level after several hours is predicted based on (calorie), there is a problem that it is difficult to improve the prediction accuracy. One possible reason for this is that the objective input information described above lacks input information necessary for predicting blood glucose levels. On the other hand, it is conceivable to add advanced information such as the above-mentioned glycohemoglobin, but since this type of information requires clinical examination, it cannot be easily obtained at the user level and is practical. is not.
The present invention is for solving such problems, and a blood sugar level predicting apparatus and a blood sugar level predicting model creating apparatus capable of improving blood sugar level predicting accuracy without requiring an advanced test technique used in a clinical test , And aims to provide a program.

  In order to achieve such an object, a blood sugar level predicting apparatus according to the present invention is a blood sugar level predicting apparatus that predicts a user's blood sugar level using a prediction model, and the amount of energy consumed or consumed by the user. , And a plurality of history data composed of a set of model input values including physical condition variable values derived from the subjective judgment results of the user regarding his / her physical condition and model output values indicating blood glucose levels corresponding to these model input values A storage unit that stores a prediction model created based on the above, and an arithmetic processing unit that predicts a blood glucose level corresponding to an arbitrary prediction condition using the prediction model of the storage unit.

  At this time, the amount of energy consumed or consumed by the user on the previous day may be used as the amount of energy, and the blood glucose level of the user measured at the early morning fasting on the next day may be used as the blood glucose level.

  A blood sugar level prediction model creation device according to the present invention is a blood sugar level prediction model creation device that creates a prediction model for predicting a user's blood sugar level, the amount of energy consumed or consumed by the user, and Generates multiple historical data consisting of a set of model input values including physical condition variable values derived from the subjective judgment results of the user regarding his / her physical condition and model output values indicating blood glucose levels corresponding to these model input values A history data generating means for generating a prediction model for predicting a user's blood glucose level based on the history data.

  At this time, the amount of energy consumed or consumed by the user on the previous day may be used as the amount of energy, and the blood glucose level of the user measured at the early morning fasting on the next day may be used as the blood glucose level.

  In addition, a program according to the present invention includes a storage unit that stores various types of information and an arithmetic processing unit that performs arithmetic processing on the information, and the computer of the information processing apparatus that predicts a user's blood glucose level using a prediction model. Model input value including physical variable values derived from the user's subjective judgment results regarding the amount of energy consumed or consumed by the user and his / her physical condition, and model output indicating blood glucose level corresponding to these model input values A step of storing a prediction model created based on a plurality of history data consisting of a pair with a value in a storage unit and a step of predicting a blood glucose level corresponding to an arbitrary prediction condition using the prediction model It is a thing.

  Another program according to the present invention includes a storage unit that stores various types of information and an arithmetic processing unit that performs arithmetic processing on the information, and performs information processing for creating a prediction model for predicting a user's blood glucose level. Model input values including the amount of energy consumed or consumed by the user and the physical condition variable values derived from the subjective judgment results of the user regarding his / her physical condition, and blood glucose levels corresponding to these model input values A history data generation step for generating a plurality of history data consisting of a pair of model output values indicating a model and a model creation step for creating a prediction model for predicting a user's blood glucose level based on the history data It is a thing.

  At this time, the amount of energy consumed or consumed by the user on the previous day may be used as the amount of energy, and the blood glucose level of the user measured at the early morning fasting on the next day may be used as the blood glucose level.

  According to the present invention, a model input value including a physical condition variable value derived from a user's subjective judgment result regarding the amount of energy acquired or consumed by the user and his / her physical condition, and a blood glucose level measured by the user are included. Multiple historical data are generated from a set of model output values, a prediction model is created based on these historical data, and a blood glucose level is predicted for an arbitrary prediction condition. The user can easily collect various model input values.

Therefore, the accuracy of predicting blood glucose level is not required without requiring advanced testing techniques as in the case of using glycohemoglobin as a new model input value, and without giving a physical or mental burden to the user. Can be improved, and the blood sugar level can be predicted with high accuracy.
As a result, not only medical institutions but also home users can predict blood glucose levels with high accuracy, and can widely support users' life management and health maintenance, as well as user life management and health by medical institutions. The burden for maintenance can be greatly reduced.

Next, embodiments of the present invention will be described with reference to the drawings.
[First Embodiment]
First, a blood glucose level prediction model creating apparatus according to a first embodiment of the present invention will be described with reference to FIG. FIG. 1 is a block diagram showing a configuration of a blood sugar level prediction model creation device according to the first embodiment of the present invention.
This blood glucose level prediction model creation device 1 is a physical condition variable derived from the user's subjective judgment results regarding his / her physical condition in addition to conventional objective input information such as the user's dietary amount, exercise amount, and blood glucose level. It is a device that uses subjective input information called values and creates a prediction model used for blood glucose level prediction from these input information.

The blood glucose level prediction model creation apparatus 1 is composed of a computer as a whole, and includes a screen display unit 11, an operation input unit 12, a data interface unit (hereinafter referred to as a data I / F unit) 13, a storage unit 14, and an arithmetic processing unit 15. Is provided.
The screen display unit 11 includes a screen display device such as an LCD or a CRT, and displays various information such as processing menus and collected data on the screen based on instructions from the arithmetic processing unit 15.
The operation input unit 12 includes an operation input device such as a keyboard and a mouse, detects a user operation, and outputs it to the arithmetic processing unit 15.
The data I / F unit 13 exchanges various data by performing data communication with the external device 4 via the communication line 3 or accessing a recording medium.

The storage unit 14 includes a storage device such as a hard disk or a memory, and stores various information and programs 14D necessary for information processing in the arithmetic processing unit 15. Typical examples of various types of information include collected data 14A, history data 14B, and prediction model 14C.
The collected data 14A is input information necessary for creating a blood glucose level prediction model, and is used for data communication with an arbitrary external device 4 such as a known blood glucose level measuring device or calorie calculator, or the operation input unit 12. It is acquired by the operation input by the user at.

The history data 14 </ b> B is information including a pair of a model input value and a model output value used for creating a blood sugar level prediction model in the collected data 14 </ b> A, and is generated by the arithmetic processing unit 15.
The prediction model 14C is a black box prediction model that estimates and outputs an output value corresponding to a given prediction input value, and is created by the arithmetic processing unit 15 based on the history data 14B.
The program 14 </ b> D is a list describing information processing procedures in the arithmetic processing unit 15. The program 14 </ b> D is fetched in advance from a communication line such as AN or a recording medium via the data I / F unit 13 and stored in the storage unit 14. .

The arithmetic processing unit 15 has a microprocessor such as a CPU and its peripheral circuits, and reads and executes the program 14D from the storage unit 14, thereby realizing various functional means by cooperating the hardware and the program 14D. It is a functional part to do.
As functional means, there are data acquisition means 15A, data processing means 15B, history data generation means 15C, and model creation means 15D.

The data acquisition unit 15A is a functional unit that takes in the collected data 5 through the data I / F unit 13 and the communication line 3 or the operation input unit 12 and stores it in the storage unit 14 as the collected data 14A.
The data processing means 15B is a functional means for processing the collected data 14A used as a model input value or model output value in order to improve the estimation accuracy of the prediction model. Examples of the processing here include moving average correction described later.

The history data generation means 15C is based on the collected data 14A and the data obtained by the data processing means 15B, and the history data 14B used to create the prediction model 14C from a set of model input values and corresponding model output values. Is a functional means for generating
The model creation unit 15D is a functional unit that creates a desired prediction model 14C that predicts and outputs a corresponding output value from a given input value based on the history data 14B.

[Collected data]
Next, the collected data 14A will be described with reference to FIGS. FIG. 2 is an example of collected data showing the amount of dinner and the amount of exercise. FIG. 3 is an example of collected data showing blood glucose levels and physical condition variables.
As the amount of meal, a commercially available calorie calculator or a calorie calculation software for a personal computer is used to input the amount of meal for each morning, noon, and evening meal and the total amount of meal per day. Depending on the user, data may be lost when they do not take 3 meals a day or when they forget to record. In that case, an average value of all data used in the analysis may be calculated and interpolated by regarding it as a missing value.

As the amount of exercise, a commercially available pedometer is used to input the number of steps per day, the total energy consumption, the amount of exercise, and the like. Small pedometers require users to take note of the number of steps every day, and forgetting to record easily occurs. Therefore, using a commercially available acceleration sensor type lifestyle recorder, the number of steps, the amount of exercise, and the energy consumption can be calculated. You may record automatically. Some devices of this type can transfer information to a personal computer with a data communication function using a dedicated program.
Since the amount of exercise has the function of lowering the blood glucose level, the value was normalized to a negative value when used as an input variable. The amount of exercise may be lost due to forgetting to wear a lifestyle recorder. Therefore, interpolation may be performed in the same manner as the amount of meal.

  The blood glucose level is measured once or several times a day using a commercially available blood glucose meter, for example, when the user is in a specific state such as early morning fasting blood glucose level (BG: Blood Glucose). Enter. In addition, if possible, if there is a total of 7 data per day during early morning fasting, before meals and 2 hours after meals, and before going to bed, approximate daily fluctuations in blood glucose levels can be found, leading to improved prediction accuracy.

The physical condition variable is an evaluation value that the user himself / herself subjectively evaluates the physical condition of the user. Diabetic patients are prone to changes in their physical condition, such as sweating, anxiety, motives, headaches, blurred vision, and hunger when they become hypoglycemic, and fever, diarrhea, vomiting, etc. Symptoms appear.
In the present embodiment, the user is asked to evaluate these physiological states occurring in the body in five stages, and when the user measures the fasting blood glucose level every morning and early in the morning, the physical condition at that time is expressed by the numeral 1 ( Self-assessment was performed with a good physical condition) to 5 (poor physical condition). Note that the number of evaluation stages is not limited to five, and the evaluation may be input with an arbitrary number of two or more.

  In addition, the evaluation variables are not limited to those directly using the evaluation values input from the user, but are known subjective evaluation value derivation techniques (for example, Kotaro Sugisu, “Stress measurement method”, (Yes ) Kawashima Shoten, see 1999/12/01 issue). For example, a checklist consisting of a plurality of questions can be presented to the user, and a physical condition variable value can be derived based on a predetermined algorithm from a result of answering the checklist, that is, a subjective judgment result of the user regarding his / her physical condition.

Normally, it is considered that the blood sugar level is relatively high when the physical condition is poor (when the physical condition variable is numerically large), and the blood sugar level is relatively low when the physical condition is low (when the physical condition variable is numerically small). Therefore, the shortage of input information is improved and prediction accuracy can be improved by creating a prediction model using physical variable values obtained by quantifying such user's subjective evaluation.
In addition, what is necessary is just to leave the evaluation criteria of a physical condition to the self-judgment for every user. Self-evaluation shall be performed before blood glucose level measurement. This is because prejudice is born by knowing one's own blood glucose level, and there is a possibility that a clear correlation may occur with the blood glucose level. By allowing self-evaluation before blood glucose measurement, a physical condition variable value that maintains the user's subjectivity can be obtained.

[Prediction model and historical data]
Next, the prediction model 14C and the history data 14B will be described with reference to FIG. FIG. 4 is an explanatory diagram showing a prediction model and history data.
In general, a prediction model is a model obtained by modeling a causal relationship or an inference process between a predicted input value serving as a prediction condition and a desired predicted output value derived from the predicted input value. As shown in FIG. 4, the prediction model 14 </ b> C handled in this embodiment has a prediction input value 6 </ b> A, such as the exercise input, the amount of meal, the physical condition variable value, and the blood glucose level obtained when the prediction input value is given. The predicted output value 6B corresponding to the value, for example, the blood glucose level in the early morning of the next day is output.

  In order to create such a prediction model 14C, history data 14B indicating the behavior of the user's living body is required. The history data 14B includes a plurality of sets 16 each including a combination of model input values and model output values acquired in advance. The model input value and the model output value of the set 16 are measured by the external device 4 from the user or input by operation and stored in the storage unit 14 as collected data 14A. Further, as necessary, the correction data of the collected data 14A by the data processing means 15B of the arithmetic processing unit 15 is used as the model input value and model output value of the set 16.

The configuration of the prediction model 14C may be a mathematical model in which the causal relationship or inference process is expressed by a mathematical formula using a physical law, or a black box prediction model in which the causal relationship or inference process is not clarified by a mathematical formula.
Among these, as black box prediction models, there are models using case bases, fuzzy inference bases, and neural networks. In particular, the case-based reasoning model is a well-known modeling technique that can be applied to a general object in which the continuity of the input / output relationship of the system is established based on the concept of topology (for example, JP-A-11-372898). (See publications).

  In general modeling techniques, it is necessary to identify model parameters such as model order and network structure, but the case-based reasoning model identifies the phase of the input space by specifying the desired output tolerance. ing. That is, the case-based reasoning model is a prediction model having case data for each partial section in the input space obtained by quantizing the model input value of each history data according to a desired output tolerance, and each case data Consists of data indicating the partial section and representative values representing model output values included in one or more history data having model input values belonging to the partial section.

  Therefore, each historical data is stored as case data in each partial section in the input space identified as described above, and at the time of output estimation, the representative value of the case data of the partial section corresponding to the model input value is the estimation result. Become. At this time, the reliability of the estimated output value can be shown by the phase distance (similarity) in the input space between the model input value and the case data stored in advance.

[Operation of blood glucose level prediction model creation device]
Next, the operation of the blood sugar level prediction model creation device according to the present embodiment will be described with reference to FIG. FIG. 5 is a flowchart showing blood sugar level prediction model creation processing according to the present embodiment.
The arithmetic processing unit 15 reads and executes the program 14D from the storage unit 14 in response to a prediction model creation instruction from the operation input unit 12 for a predetermined processing menu displayed on the screen display unit 11, thereby executing the blood glucose of FIG. The value prediction model creation process starts.

  Here, a case will be described in which a case-based prediction model is created in which a predicted output value consisting of an early morning blood glucose level on the day i is predicted from a predicted input value obtained on the previous day i-1. At this time, with respect to the model input value, as the amount of energy consumed / consumed by the user, the amount of dinner consumed by the user on the previous day i-1 (calorie intake: Qin) and the amount of exercise performed by the user on the previous day i-1. (Calorie consumption: Qout) and using a physical condition variable value (PC) indicating the physical condition of the self evaluated by the user in the early morning fasting on the previous day i-1 and a blood glucose level (BG ′) measured at that time, a model The case where the blood glucose level (BG) measured at the early morning fasting of i on the day is used as an output value will be described as an example.

  First, the arithmetic processing unit 15 takes in the collected data 5 from the external device 4 via the data I / F unit 13 by the data acquisition unit 15A, and stores it in the storage unit 14 as the collected data 14A (step 100). The acquisition step of the collected data 14A may be stored in the storage unit 14 in advance prior to the prediction model creation process.

Next, the arithmetic processing unit 15 reads out the blood glucose level used as the model output value from the collected data 14A in the storage unit 14 by the data processing unit 15B, and performs the moving averaging process (step 101).
As a specific example of this moving averaging process, in formula (1) of FIG. 5, seven blood glucose levels BGi−6 to blood glucose level BGi obtained from six days before to the measurement date i on the basis of the measurement date i. By obtaining the average value, the blood-averaged blood glucose level BGmi on the measurement date i is obtained. Note that the blood glucose level BGm ′ that has been moving averaged in the same manner as described above may also be used for the blood glucose level BG ′.

Then, the arithmetic processing unit 15 uses the history data generation unit 15C to measure each model input data of the i-1 day before measurement date used as a model input value, that is, Qin, Qout, PC, and BGm ′, and a measurement used as a model output value. The model output data BGm of day i is combined for each measurement date to generate individual history data 14B and store it in the storage unit 14 (step 102).
Subsequently, the calculation processing unit 15 reads out the history data 14B from the storage unit 14 by the model creation unit 15D, and estimates and outputs a corresponding output value from a given input value using a predetermined algorithm. 14C is created (steps 103 and 104).

As an algorithm for creating a prediction model, an algorithm corresponding to the inference model is used. For example, in a mathematical model, the solution of a linear function or nonlinear function with each model input data used as a model input value as an input variable and the model output data used as a model output value as an output variable is derived based on each history data. A known algorithm is used.
In addition, in the black box prediction model using the neural network, the process of learning each history data is repeatedly executed. In the model using the fuzzy inference base, the process of generating individual rules from each history data. Will be executed repeatedly.

Of the black box prediction models, a model using a case base creates a prediction model by the following algorithm. FIG. 6 is an explanatory diagram showing generation processing of case data constituting the prediction model, and shows a case where the model input value is composed of two input variables x1 and x2, and the model output value is composed of y.
First, an input space x1-x2 made up of model input data used as model input values is divided into a grid, and a plurality of partial sections called meshes are provided. At this time, the size of the mesh is determined so that the variation width of the model output value of each history data having the model input value belonging to each mesh is equal to or smaller than the allowable output error.

After each history data 14B is distributed to each mesh, a mesh having one or more history data is selected as an example from these meshes, and input / output values and output values representing the mesh are calculated from the history data. And output as case data.
For example, as shown in FIG. 6, when three pieces of history data are distributed to the same mesh, these are integrated as one case. At this time, the average value of the outputs y of the three history data is used as the output value representing the case, and the median value of the mesh is used as the input value representing the case.

  In this way, case data is generated from each history data 14B (FIG. 5: Step 103), the prediction model 14C consisting of these case data is stored in the storage unit 14 (FIG. 5: Step 104), and a series of predictions are performed. The model creation process ends.

  FIG. 7 shows a prediction result by the prediction model created by the blood sugar level prediction model creation device according to the present embodiment. Here, a change in blood glucose level obtained by the prediction model, for example, an increase / decrease relative to the blood glucose level measured last time is determined, and a coincidence rate with the change in the actually measured blood glucose level is calculated. Further, as a comparison target, a similar blood glucose level change coincidence rate was calculated using a prediction model that does not use a physical condition variable value.

Each of the prediction models A to E has a different combination of model input values used for creating the prediction model. The prediction model A was created using the dinner amount on the previous day, the total exercise amount on the previous day, the physical condition variable value on the previous day, and the blood glucose level on the previous day as model input values.
The prediction model B was created using the dinner volume on the previous day, the physical condition variable value on the previous day, and the blood glucose level on the previous day as model input values.

The prediction model C was created using the total meal amount on the previous day, the physical condition variable value on the previous day, and the blood glucose level on the previous day as model input values.
The prediction model D was created using the total exercise amount on the previous day, the physical condition variable value on the previous day, and the blood glucose level on the previous day as model input values.
The prediction model E was created using the total meal amount on the previous day, the total exercise amount on the previous day, the physical condition variable value on the previous day, and the blood glucose level on the previous day as model input values.

  As shown in FIG. 7, the prediction model according to the present embodiment using the physical condition variable value is higher in the blood glucose level change coincidence rate than the conventional prediction model not using the physical condition variable value. It was. Moreover, the same improvement tendency was seen also in any prediction model AE.

  As described above, in the present embodiment, the model input value including the physical condition variable value derived from the user's subjective judgment result regarding the amount of energy acquired or consumed by the user and the user's physical condition, and measurement from the user are performed. Since a plurality of historical data is generated from a set of model output values consisting of blood glucose levels, a prediction model for predicting a user's blood glucose level as a desired output value from a new input value is created based on these historical data. The user can easily input model input values necessary for creating the prediction model.

Therefore, the accuracy of predicting blood glucose level is not required without requiring advanced testing techniques as in the case of using glycohemoglobin as a new model input value, and without giving a physical or mental burden to the user. Can be improved.
As a result, not only medical institutions but also home users can predict blood glucose levels with high accuracy, and can widely support users' life management and health maintenance, as well as user life management and health by medical institutions. The burden for maintenance can be greatly reduced.

  In addition, as the model input value, a physical condition variable value composed of an evaluation value that the user himself / herself subjectively evaluated the physical condition of the user is used, so that the physiological state of the user obtained using advanced examination technology is used. It is possible to create a prediction model that takes into account the user's overall physiological state, that is, the user's physical condition and physiological changes, which are difficult to obtain even using advanced examination techniques, and can greatly improve blood glucose level prediction accuracy.

Also, as the amount of energy, the amount of energy consumed or consumed by the user the previous day is used, and the blood glucose level of the user measured at the early morning hunger the next day is used as the blood glucose level. The blood sugar level of the next day can be predicted.
As a result, it is possible to predict a significant change in blood glucose level particularly in the middle of the night or early morning, and to take measures to eliminate risk factors in the diabetic patient caused by the blood glucose level change in advance. Can be implemented appropriately.

[Second Embodiment]
Next, a blood glucose level prediction apparatus according to the second embodiment of the present invention will be described with reference to FIG. FIG. 8 is a block diagram showing a configuration of a blood sugar level predicting apparatus according to the second embodiment of the present invention.
This blood glucose level prediction device 2 uses a prediction model created by the blood glucose level prediction model creation device according to the first embodiment described above, and uses a desired blood glucose level as a predicted output value corresponding to a predetermined predicted input value. It is a device that predicts.

The blood glucose level prediction device 2 is composed of a computer as a whole, and includes a screen display unit 21, an operation input unit 22, a data interface unit (hereinafter referred to as a data I / F unit) 23, a storage unit 24, and an arithmetic processing unit 25. ing.
The screen display unit 21 includes a screen display device such as an LCD or a CRT, and displays various information such as a processing menu, a predicted input value, or a prediction result on the screen based on an instruction from the arithmetic processing unit 25.
The operation input unit 22 includes an operation input device such as a keyboard and a mouse, detects a user operation, and outputs the operation to the arithmetic processing unit 25.
The data I / F unit 23 exchanges various data by performing data communication with an external device via a communication line.

The storage unit 24 includes a storage device such as a hard disk or a memory, and stores various information and programs 24C necessary for information processing in the arithmetic processing unit 25. Typical information includes a prediction model 24A and a prediction input value 24B.
The prediction model 24A is the prediction model 14C created by the blood sugar level prediction model creation device 1 according to the first embodiment described above.

The predicted input value 24B is information indicating an input condition necessary for blood glucose level prediction, and is acquired via the operation input unit 22 and the data I / F unit 23.
The program 24C is a list describing information processing procedures in the arithmetic processing unit 25. The program 24C is fetched in advance from a communication line such as a LAN or a recording medium via the data I / F unit 23 and stored in the storage unit 24. .

The arithmetic processing unit 25 has a microprocessor such as a CPU and its peripheral circuits, and reads and executes the program 24C from the storage unit 24, thereby realizing various functional means by cooperating the hardware and the program 24C. It is a functional part to do.
Functional means include data acquisition means 25A, prediction means 25B, and output means 25C.

The data acquisition unit 25A is a functional unit that captures the collected data 6 via the data I / F unit 23 and the communication line 3 or the operation input unit 22, and stores the collected data 6 as the predicted input value 24B in the storage unit 24. .
The prediction unit 25B is a functional unit that predicts a blood glucose level corresponding to the predicted input value 24B by referring to the prediction model 24A of the storage unit 24 based on the predicted input value 24B of the storage unit 24.
The output unit 25C is a functional unit that compares the blood glucose level obtained by the predicting unit 25B with the blood glucose level measured last time and outputs the change, that is, the increase / decrease as a prediction result.

[Operation of blood glucose level prediction device]
Next, with reference to FIG. 9, the operation of the blood sugar level predicting apparatus according to the present embodiment will be described. FIG. 9 is a flowchart showing blood sugar level prediction processing according to the present embodiment.
The arithmetic processing unit 25 reads and executes the program 24C from the storage unit 24 in response to a prediction model creation instruction from the operation input unit 22 for a predetermined processing menu displayed on the screen display unit 21, thereby executing the blood glucose of FIG. The value prediction process starts.

  First, the arithmetic processing unit 25 uses the data acquisition unit 25A to capture a predicted input value serving as a prediction condition via the operation input unit 22, and stores it as a predicted input value 24B in the storage unit 24 (step 200). The step of capturing the predicted input value 24B may be stored in the storage unit 24 in advance prior to the blood glucose level prediction process. It is assumed that the prediction model 24A is stored in the storage unit 24 in advance prior to the blood sugar level prediction process.

Next, the arithmetic processing unit 25 reads the prediction input value 24B and the prediction model 24A from the storage unit 24 by the prediction unit 25B, refers to the prediction model 24A based on the prediction input value 24B (step 201), and performs prediction. A blood glucose level corresponding to the input value 24B is derived (step 202).
FIG. 10 is an explanatory diagram showing a derivation process of a predicted output value using a case-based prediction model, in which the predicted input value is composed of two input variables x1 and x2, and the predicted output value is composed of y. Yes.

  As described above, the case-based prediction model includes a plurality of case data corresponding to meshes provided in the input space. Here, when the predicted input value A is given, a case corresponding to the predicted input value A is searched. Then, the predicted output value y of the retrieved case is derived as an output estimated value, that is, a predicted blood glucose level.

When there is no case corresponding to the given predicted input value A, the predicted blood glucose level is estimated using a concept called similarity as shown in FIG.
The similarity is the degree of similarity between each mesh provided in the input space of the prediction model 24A (case base) and the mesh corresponding to the new prediction condition, that is, the input data. It is a scale to show. As shown in FIG. 10, the case where a case exists in the center mesh corresponding to the predicted input value A is defined as similarity = 0, and the case existing next to the center mesh is “similarity = 1”. Thereafter, the degree of similarity increases by 1 each time the mesh is moved away from the center mesh.

In the case of FIG. 11, since a case exists in the mesh next to the center mesh, similarity = 1. In this case, the average value of the predicted output value y of these cases is obtained for the cases of the meshes of similarity = 1 existing around the central mesh, and this average value is obtained as the estimated output value of the central mesh, that is, the predicted blood glucose level. Derived.
Therefore, the estimated value based on the case of similarity i has accuracy within (i + 1) × allowable output. At this time, when the examples on both sides are used successfully with respect to the central mesh, it is expected that the output value has an accuracy better than (i + 1) × output allowable width. Further, when only one case is used for the central mesh, it is expected based on the continuity of input and output that the accuracy is about (i + 1) × allowable output width.

  In the arithmetic processing unit 25, after the predicted output value corresponding to the predicted input value is derived by the predicting unit 25B as described above, the predicted blood glucose level derived as the predicted output value by the output unit 25C is compared with a predetermined comparison value. A reference, for example, the blood glucose level measured the previous day (or predicted the previous day) is compared. Then, the change, that is, the increase / decrease is output as a prediction result (step 203), and the series of blood glucose level prediction processes is terminated.

  As described above, in the present embodiment, the model input value including the physical condition variable value derived from the user's subjective judgment result regarding the amount of energy acquired or consumed by the user and the user's physical condition, and measurement from the user are performed. Since the prediction model generated based on the history data consisting of the model output value consisting of the blood glucose level is used to predict the blood glucose level corresponding to the newly input prediction input value, The user can easily input the necessary predicted input value.

Therefore, the accuracy of predicting blood glucose level is not required without requiring advanced testing techniques as in the case of using glycohemoglobin as a new model input value, and without giving a physical or mental burden to the user. Can be improved.
As a result, not only medical institutions but also home users can predict blood glucose levels with high accuracy, and can widely support users' life management and health maintenance, as well as user life management and health by medical institutions. The burden for maintenance can be greatly reduced.

  In addition, as the model input value, a physical condition variable value composed of an evaluation value that the user himself / herself subjectively evaluated the physical condition of the user is used, so that the physiological state of the user obtained using advanced examination technology is used. It is possible to create a prediction model that takes into account the user's overall physiological state, that is, the user's physical condition and physiological changes, which are difficult to obtain even using advanced examination techniques, and can greatly improve blood glucose level prediction accuracy.

Also, as the amount of energy, the amount of energy consumed or consumed by the user the previous day is used, and the blood glucose level of the user measured at the early morning hunger the next day is used as the blood glucose level. The blood sugar level of the next day can be predicted.
As a result, it is possible to predict a significant change in blood glucose level particularly in the middle of the night or early morning, and to take measures to eliminate risk factors in the diabetic patient caused by the blood glucose level change in advance. Can be implemented appropriately.

[Extended for each embodiment]
In each of the above embodiments, the blood sugar level prediction model creation device 1 and the blood sugar level prediction device 2 have been described separately, but may be realized by combining them into one device. For example, the data processing unit 15B, the history data generation unit 15C, and the model creation unit 15D in the calculation processing unit 15 of the blood sugar level prediction model creation device 1 may be realized by the calculation processing unit 25 of the blood sugar level prediction device 2. At this time, the blood glucose level (actually measured value) of the user obtained sequentially is input to the blood glucose level prediction device 2 as a model output value, and the adaptive learning process of the prediction model 14C may be performed sequentially or as necessary. .

  Moreover, although each embodiment mentioned above demonstrated as an example the case where the prediction model 14C was created by making into a model input value the exercise amount obtained from the user's step count, and the amount of meal which the user ingested, The amount of exercise and the amount of meal are one specific index indicating the amount of energy acquired or consumed by the user. Therefore, as the model input value used for creating the blood sugar level prediction model, any other index may be used as long as it is an index indicating the amount of energy acquired or consumed by the user.

  Further, the acquired energy amount and the consumed energy amount of these users have a certain degree of correlation, as can be seen from FIG. 2 described above. Therefore, although both of these energy amounts and physical condition variable values may be used in combination as model input values, any one of the energy amounts and physical condition variable values may be used in combination.

  Further, in each of the above embodiments, as an example of predicting the early morning fasting blood glucose level on the next day based on the predicted input value obtained on the previous day, the prediction model is created and the blood glucose level using the prediction model is calculated. Although the prediction has been described, the prediction timing is not limited to this. For example, when predicting a blood glucose level with a shorter span such as several hours later, or when predicting a blood glucose level with a longer span such as after 24 hours, a prediction model is created in the same manner as described above. Can be used to predict the desired blood glucose level.

  Further, the time when each model input value or predicted input value is acquired is not limited to the previous day as described above. For example, the time position of a desired blood glucose level to be predicted, such as a physical condition variable value obtained on that day, What is necessary is just to select timely according to prediction accuracy.

It is a block diagram which shows the structure of the blood glucose level prediction model creation apparatus concerning the 1st Embodiment of this invention. It is an example of collection data which shows the amount of dinner and the amount of exercise. It is an example of the collection data which shows a blood glucose level and a physical condition variable. It is explanatory drawing which shows a prediction model and historical data. It is a flowchart which shows the blood glucose level prediction model creation process concerning the 1st Embodiment of this invention. It is explanatory drawing which shows the production | generation process of the case data which comprises a prediction model. It is the prediction result by the prediction model created with the blood glucose level prediction model creation apparatus concerning the 1st Embodiment of this invention. It is a block diagram which shows the structure of the blood glucose level prediction apparatus concerning the 2nd Embodiment of this invention. It is a flowchart which shows the blood glucose level prediction process concerning the 2nd Embodiment of this invention. It is explanatory drawing which shows the derivation | leading-out process (similarity = 0) of the prediction output value by a case-based prediction model. It is explanatory drawing which shows the derivation | leading-out process (similarity = 1) of the prediction output value by a case-based prediction model.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Blood glucose level prediction model preparation apparatus, 11 ... Screen display part, 12 ... Operation input part, 13 ... Data I / F part, 14 ... Memory | storage part, 14A ... Collected data, 14B ... History data, 14C ... Prediction model, 14D DESCRIPTION OF SYMBOLS ... Program, 15 ... Arithmetic processing part, 15A ... Data acquisition means, 15B ... Data processing means, 15C ... History data generation means, 15D ... Model creation means, 2 ... Blood glucose level prediction apparatus, 21 ... Screen display part, 22 ... Operation Input unit 23 ... Data I / F unit 24 ... Storage unit 24A ... Prediction model 24B ... Predicted input value 24C ... Program 25 ... Operation processing unit 25A ... Data acquisition unit 25B ... Prediction unit 25C ... Output means, 3 ... communication line, 4 ... external device, 5 ... collected data, 6A ... predicted input value, 6B ... predicted output value.

Claims (7)

A blood glucose level prediction device that predicts a user's blood glucose level using a prediction model,
Model input values including the amount of energy consumed or consumed by the user and physical condition variable values derived from the subjective judgment results of the user regarding his / her physical condition, and model output values indicating blood glucose levels corresponding to these model input values A storage unit for storing a prediction model created based on a plurality of historical data consisting of
A blood glucose level prediction apparatus comprising: an arithmetic processing unit that predicts a blood glucose level corresponding to an arbitrary prediction condition using the prediction model of the storage unit. In the blood glucose level prediction apparatus according to claim 1,
The amount of energy is composed of the amount of energy consumed or consumed by the user on the previous day, and the blood glucose level is composed of the blood glucose level of the user measured in the early morning of the next day. . A blood glucose level prediction model creation device for creating a prediction model for predicting a user's blood glucose level,
Model input values including the amount of energy consumed or consumed by the user and physical condition variable values derived from the subjective judgment results of the user regarding his / her physical condition, and model output values indicating blood glucose levels corresponding to these model input values History data generating means for generating a plurality of history data consisting of
A blood glucose level prediction model creating apparatus comprising: a model creating unit that creates a prediction model for predicting a user's blood glucose level based on the history data. In the blood sugar level prediction model creation device according to claim 3,
The energy amount is an amount of energy consumed or consumed by the user on the previous day, and the blood glucose level is the blood glucose level of the user measured on the next day at the early morning hungry level. Creation device. A computer of an information processing apparatus that has a storage unit that stores various types of information and an arithmetic processing unit that performs arithmetic processing on the information, and predicts a user's blood glucose level using a prediction model
Model input values including the amount of energy consumed or consumed by the user and physical condition variable values derived from the subjective judgment results of the user regarding his / her physical condition, and model output values indicating blood glucose levels corresponding to these model input values Storing in the storage unit a prediction model created based on a plurality of historical data consisting of
The program which performs the step which estimates the blood glucose level corresponding to arbitrary prediction conditions using the said prediction model. A computer of an information processing apparatus that has a storage unit that stores various types of information and an arithmetic processing unit that performs arithmetic processing on the information, and creates a prediction model for predicting a user's blood glucose level.
Model input values including the amount of energy consumed or consumed by the user and physical condition variable values derived from the subjective judgment results of the user regarding his / her physical condition, and model output values indicating blood glucose levels corresponding to these model input values A history data generation step for generating a plurality of history data consisting of
A program for executing a model creation step of creating a prediction model for predicting a blood glucose level of a user based on the history data. In the program according to claim 5 or 6,
The program is characterized in that the amount of energy is the amount of energy consumed or consumed by the user on the previous day, and the blood glucose level is the blood glucose level of the user measured on the next day on an early morning hungry day.

Images