JP2020031701A - Event estimation method, event estimation program, server device, biological information measuring device, and event estimation system - Google Patents

Abstract

To provide an event estimation method, an event estimation program, a server device, biological information measuring device, and biological information measuring system capable of improving a user's convenience in event input processing.SOLUTION: An event estimation method includes: a first step for receiving first blood sugar data and an event content related to a measuring date of the first blood sugar data for a predetermined period; a second step for calculating time-series variation parameters of the first blood sugar data, analyzing relationships between the first blood sugar data and the event content by machine learning, and generating a classification model; and a third step for inputting the time-series variation parameters of the blood sugar for a predetermined period on received second blood sugar data in the classification model, and estimating an event content related to a measuring date of the second blood sugar data.SELECTED DRAWING: Figure 21

Description

  The present invention relates to, for example, an event estimation method, an event estimation program and a server device, a biological information measurement device, and an event estimation system displayed on a biological sensor device for measuring the concentration of sugars and amino acids in a living body.

  The configuration of the conventional biological information measurement device includes an acquisition unit that captures biological information from a sensor, a display unit that displays the biological information captured by the acquisition unit, and an input unit that inputs an event that affects a change in blood glucose level. (For example, see Patent Document 1).

  Further, a system for more accurately predicting a blood glucose level according to the type of an input event (for example, exercise, meal, insulin administration, etc.) is disclosed (for example, see Patent Document 2).

JP 2011-206486 A JP 2012081166 A

The problem of the above-mentioned conventional biological information measuring device is that the convenience for the user is low.
That is, the conventional biological information measuring device is configured to input the state (event) of the user during the measurement period of the blood sugar level. However, the user has to input, for each event, an event that affects fluctuations in blood sugar level such as exercise, diet, insulin administration, etc., so that the user is not very convenient. Was.

  Therefore, an object of the present invention is to improve the convenience of a biological information measuring device with respect to an event input process.

  In order to achieve this object, the event estimation method of the present invention provides a first blood glucose level data and a first blood glucose level data, the event content relating to the measurement date and time of the first blood glucose level data being received for a predetermined period. A second step of calculating a time-series variation parameter of the first blood glucose level data, analyzing the relationship between the first blood glucose level data and the event content by machine learning, and generating a classification model; A parameter of time-series variation of blood glucose level for a predetermined period of the received second blood glucose level data is input, and a classification model is used to estimate an event content related to the date and time of measurement of the second blood glucose level data. 3 steps.

  The present invention can improve the convenience of the user of the biological information measuring device, for example.

The figure which shows the use condition of the continuous blood glucose measurement (CGM) apparatus which provides the blood glucose level information to the event estimation system which concerns on one Embodiment of this invention. The figure which shows the use condition of the continuous blood glucose measurement (CGM) apparatus of FIG. FIG. 2 is a control block diagram of an electronic unit included in the continuous blood glucose measurement (CGM) device of FIG. 1. FIG. 4 is a view showing an example of a display screen of the terminal device according to the embodiment. The figure which shows an event content and numerical information. The figure which shows the structure of the event estimation system by monitoring a blood glucose level. The figure which shows a classification model. FIG. 2 is a diagram showing an example of a configuration of a classification model (here, a neural network) of the embodiment. The figure which shows a regression model. FIG. 3 is a diagram illustrating an example of a configuration of a regression model (here, logistic regression) of the embodiment. FIG. 2 is a control block diagram showing a simple configuration of the event estimation system. 12 is a flowchart showing the flow of processing in the event estimation system in FIG. FIG. 2 is a control block diagram showing a simple configuration of the event estimation system. 14 is a flowchart showing the flow of processing in the event estimation system of FIG. 9 is a flowchart showing the flow of processing in an analysis unit on the server device side. 9 is a flowchart showing the flow of processing in the adjustment unit on the terminal device side. FIG. 3 is a control block diagram of a server device included in the event estimation system. FIG. 3 is a control block diagram of a terminal device included in the event estimation system. 4A to 4D are diagrams illustrating an example of an event input notification function of the terminal device. 4A and 4B are diagrams illustrating an example of an event input notification function of a terminal device. 5 is a flowchart of an event input notification function of the terminal device. The control block diagram of a terminal device. 4A to 4C are diagrams illustrating an example of an event automatic input notification function of a terminal device. (A), (b) is a figure in which an estimation part estimates occurrence of a drug treatment (medication) event, and shows automatic input. 9 is a flowchart showing a flow of processing in an event automatic input notification function of the terminal device. The figure which shows an example of a robot type external device.

  An event estimation method, an event estimation program, a server device, a biological information measurement device, and an event estimation system according to an embodiment of the present invention will be described below with reference to the drawings.

<Configuration of continuous blood glucose meter (CGM) device>
Hereinafter, an embodiment of the present invention applied to a glucose sensor for measuring glucose as an example of the continuous blood glucose measurement (CGM) system will be described with reference to the accompanying drawings.

The continuous blood glucose measurement (CGM) system is a system that includes the human body-mounted sensor device 1 and continuously measures a blood glucose level for a diabetic patient.
In the following description, “down” means the puncture side (the side where the needle projects) in the puncture direction in which puncture is performed on the patient, and “up” means the side opposite to the puncture side. (Puncture knob side).

<Overview of human-mounted sensor device>
FIG. 1 shows a human-worn sensor device 1 included in a continuous blood glucose measurement (CGM) system according to the present embodiment.

  The human body-mounted sensor device 1 of the continuous blood glucose measurement (CGM) system places a sensor under the skin of the upper arm 2 of a diabetic patient. Then, the glucose concentration in the interstitial fluid of the subcutaneous tissue is continuously measured.

  The human body-mounted sensor device 1 according to the present embodiment converts the glucose concentration into a current value, calculates the current value, and transmits the value. For example, the human body-mounted sensor device 1 measures a current value every 10 seconds, calculates an average of a predetermined number of times, and records the value in a memory. Then, for example, the current value for each minute is stored by storing the value obtained by calculating the average of the values of the six samples acquired every 10 seconds in the memory.

As shown in FIG. 1, the human body-mounted sensor device 1 communicates with a terminal device (biological information measuring device) 4 via a substantially elliptical electronic unit 3 via radio.
In the human body-mounted sensor device 1, the electronic unit 3 stores the current value from the sensor in the memory, and transmits the value stored in the memory to the terminal device 4 wirelessly. The transmitted value may be a current value or a value after calculating a blood glucose level from the current value. When the read value is the current value, the terminal device 4 displays the blood glucose value calculated from the current value together with the time information, and stores the time information and the blood glucose value in a memory in the terminal device 4.

  By continuing the measurement of the blood glucose level for about 3 to 14 days, it is possible to grasp the fluctuation of the blood glucose level of the diabetic patient over 24 hours. This enables more appropriate treatment according to the symptoms of each diabetic patient.

Further, using the blood sugar level fluctuation information, it is also possible to calculate an appropriate amount and timing of insulin to be administered to the patient.
The insulin pump administers an appropriate amount of insulin while monitoring the patient's blood glucose level in real time while being wirelessly connected to the human body-mounted sensor device 1 (and / or the terminal device 4). As a result, the function of the artificial pancreas can be realized, and ideal blood sugar level control becomes possible.

FIG. 2 is a cross-sectional view of the upper arm 2 of a diabetic patient to which the human body-mounted sensor device 1 is mounted.
A sensor 5 protrudes below the human body-mounted sensor device 1. In a state where the sensor 5 is worn on the human body, the sensor 5 is placed in a subcutaneous state.

  The tip of the sensor 5 is needle-shaped or rod-shaped, and its length is about 1 cm in order to make it easier to insert it under the skin of a diabetic patient. The tip portion of the sensor 5 is covered with a protective layer that transmits and absorbs glucose, which is a test substance, and reacts glucose with an enzyme to generate a test substance, a reaction layer including an enzyme layer, a mediator layer, and the like.

  An electrode for electrochemically measuring glucose is provided below the reaction layer. By piercing the tip of the sensor 5 subcutaneously and placing it in the interstitial fluid, the glucose concentration in the interstitial fluid of the subcutaneous tissue can be observed (and / or measured / detected).

  In such a subcutaneous indwelling glucose sensor that measures glucose in subcutaneous tissue, a time lag may occur between the blood glucose level in blood and the measurement result. Therefore, correction may be required using the measurement value of the glucose sensor of the self blood glucose measurement (SMBG).

  When mounting the human body-mounted sensor device 1 on a human body, the mount unit 6 is arranged on the skin. The sensor 5 is mounted on the mount unit 6 via a connection terminal of the sensor. The electronic unit 3 is mounted on the mount unit 6, and the electronic unit 3 and the sensor 5 are connected via the mount unit 6.

<Description of electronic unit>
FIG. 3 shows a control block diagram of the electronic unit 3.
The sensor unit 7 in FIG. 3 is a unit having the sensor 5 and the connection terminal of the sensor 5.

  The electronic unit 3 measures the current value measured by the sensor unit 7 with the measuring unit 8 and sends the measured current value to the control unit 9. The control unit 9 measures the temperature in the vicinity of the sensor unit 7 with the temperature sensor 10, corrects the temperature, and calculates the blood glucose level from the current value.

In the control section 9, the calculation of the blood sugar level is executed at predetermined time intervals. In the present embodiment, the predetermined interval is set every minute.
The control unit 9 transmits the value stored in the storage unit 11 to the terminal device 4 via the communication unit 12 according to an instruction from the terminal device 4.

  The electronic unit 3 in the present embodiment has a built-in battery 13. When the remaining battery power of the battery 13 becomes insufficient, the electronic unit 3 is discarded. The remaining battery level of the battery 13 is monitored through the terminal device 4.

  When the sensor unit 7 is replaced, the terminal device 4 determines whether or not the remaining amount of the battery 13 is sufficient before the next replacement of the sensor unit 7 with respect to the remaining battery amount of the battery 13 of the electronic unit 3. Check if.

  Here, if the remaining amount is not sufficient, the terminal device 4 instructs the user to replace the electronic unit 3 and displays the electronic unit 3 as unusable.

<Display screen of terminal device>
FIG. 4 shows an example of a display screen of the terminal device (biological information measuring device) 4 in the present embodiment.

  The layout of the display screen of the terminal device 4 in FIG. 4 is divided into a plurality. The measurement date of the blood glucose level is displayed on the first split screen 14a from the upper part of the display screen. Buttons for changing the date and time are provided on the left and right of the first split screen 14a. When this button is touched with a finger, the measurement date is changed.

When the right button is touched on the first split screen 14a, the measurement date can be advanced. By touching the left button, the measurement date can be returned to the past.
The measurement time is displayed on the left side of the second split screen 14b. On the right side, the blood glucose measurement is displayed along with the trend. In the trend, if the displayed blood glucose measurement values are viewed in chronological order, an upward-sloping arrow is displayed if the blood-sugar measurement value is rising, and a downward-sloping arrow is displayed if the displayed blood sugar measurement value is falling.

  On the third divided screen 14c, blood glucose measurement values are displayed in a graph within a predetermined time range. In the graph, the vertical axis represents the blood glucose level, and the horizontal axis represents the measurement time (time). In the graph, markers for the measurement time on the horizontal axis are displayed by vertical dotted lines. By touching this marker and moving it in the horizontal direction, the position of the marker can be made movable. The position of the marker is linked to the blood glucose measurement value displayed on the second split screen 14b, and the blood glucose measurement value corresponding to the marker position is displayed.

The fourth divided screen 14d, the fifth divided screen 14e, the sixth divided screen 14f, and the seventh divided screen 14g are screens for inputting and displaying events associated with blood glucose measurement.
An event indicates an event or state of a measurer at the time of measuring a blood sugar level. Specifically, the event content is defined as, for example, hypoglycemia awareness, sick day, menstrual day, wake-up, exercise, bathing, meal, bedtime, insulin administration, medication, and the like.

A button for inputting this event is displayed on the fourth split screen 14d.
The fifth divided screen 14e, the sixth divided screen 14f, and the seventh divided screen 14g are screens for displaying events. The event is displayed by being divided into event content and numerical information related to the event content.

FIG. 5 shows the event content of the event and numerical information associated with the event content.
The event content includes, for example, awareness of hypoglycemia, sick day, menstrual date, wake-up, exercise, bathing, meal, bedtime, insulin administration, medication, and the like. Further, as the numerical information, in the case of exercise, there are exercise duration, load, and the like. For meals, there is a carb count. For insulin administration, there is an insulin dose. And for medication, there is a dosage.

A user such as a diabetic patient needs to input the state at the time of measurement to the terminal device 4 as event content and event numerical value information while measuring the blood glucose level. However, this input operation is troublesome for the user and is inconvenient.
Therefore, in the event estimation method of the present embodiment, an event content corresponding to the blood glucose level and the event numerical value information are estimated while observing the measured blood glucose level.

<Overview of the event estimation system>
FIG. 6 shows the configuration of the blood glucose level event estimation system.
In a database 20 such as a cloud, blood sugar level data of a diabetic patient, event contents accompanying the blood sugar level data, and numerical information thereof are input.
The combination of the blood glucose level data, the event content, and the event numerical information is a combination of highly accurate data, and is preferably a large number. In university hospitals and specialized institutions, a combination of continuous blood glucose level data and event details at the measurement date and time is used.

  The blood sugar level data of these diabetic patients, the contents of the event accompanying the blood sugar level data, and the numerical information thereof are transmitted from the database 20 to the server device 21 using a network or the like.

  The server device 21 calculates various parameters of the time-series fluctuation of the blood glucose level based on the above data, and analyzes the relationship between the blood glucose level fluctuation and the event content by machine learning to generate a classification model. For this machine learning, a model that can perform a plurality of classifications such as a neural network is used.

  Further, a regression model is generated by analyzing the relationship between the blood sugar fluctuation and the numerical value by machine learning, particularly for the items of the event that require numerical input. For this machine learning, a model capable of outputting numerical values such as logistic regression is used.

  Thus, the classification model and the regression model generated by the server device 21 are transmitted to the terminal device 4 using a network or the like. The terminal device 4 performs the event estimation using the classification model and the regression model as the learned models. More specifically, event contents and numerical value information of the event are estimated for blood glucose level data of a specific user measured by mounting the human body-mounted sensor device 1 of the continuous blood glucose measurement (CGM) system.

  In addition, blood glucose level data of a specific user measured by mounting the human body-mounted sensor device 1 of the continuous blood glucose measurement (CGM) system is input to the terminal device 4. Then, the specific user inputs the content of the event accompanying the blood glucose level data and the numerical information thereof to the terminal device 4.

The blood glucose level data of the specific user, the contents of the event accompanying the blood glucose level data, and the numerical information thereof are stored in the terminal device 4.
The terminal device 4 calculates various parameters of the time series fluctuation of the blood glucose level, and further updates the classification model from the relationship between the blood glucose level fluctuation and the event content. Further, the regression model is updated with the blood glucose level data of a specific user, the contents of events accompanying the blood glucose level data, and numerical information thereof.

  With such a configuration, the event content can be estimated using the classification model and the regression model with the measured blood sugar level data as an input. As a result, it is not necessary for the user to input the contents of the event himself, so that, for example, the convenience of the user of the biological information measuring device can be improved.

  In the present embodiment, the server device 21 performs machine learning to generate a classification model and a regression model, and the terminal device 4 estimates the event content using the classification model and the regression model. did. However, the present invention is not limited to this configuration.

  For example, a configuration is also possible in which the measured blood sugar level data is input into the server device 21, the event content is estimated using the classification model and the regression model, and the estimated event content is transmitted to the terminal device 4. Good.

In addition, the model is updated when a certain number of data input by the terminal device 4 are collected and transmitted to the server device 21 and the server device 21 updates the classification model and the regression model by using these data, The updated model may be transmitted to the device 4.
Furthermore, the terminal device 4 itself may generate a classification model and a regression model by machine learning, and estimate the event content using the classification model and the regression model.

<Description of classification model>
FIG. 7 shows a classification model.

The classification model is a learning model that inputs parameters of time-series fluctuations of measured blood glucose level data, and outputs the most likely event content from the input.
The parameters of the time-series fluctuation of the blood glucose data include the gradient of the blood glucose fluctuation, the difference between the peak and the blood glucose level at the start of the blood glucose fluctuation, the time until the peak of the blood glucose fluctuation, the difference between the slope of the blood glucose fluctuation until then, and the start of the fluctuation. There are a blood sugar level at the time point, a blood sugar level at the end of the fluctuation, and the like.

FIG. 8 is a diagram illustrating an example of a configuration of a classification model (here, a neural network) of the present embodiment.
The classification model according to the present embodiment has, as an input layer, a plurality of input units for inputting parameters of time-series variation of measured blood glucose level data.

The plurality of input units X _{1 to} X _n are respectively used as parameters of the time series variation of the measured blood glucose level data, as a gradient of the blood glucose variation, a difference between the peak and the blood glucose level at the start of the blood glucose variation, and a time until the peak of the blood glucose variation. , The difference between the slope and the blood glucose fluctuation up to that time, the blood glucose level at the start of the fluctuation, and the blood glucose level at the end of the fluctuation.

  Other configurations of the neural network are the same as those of the known technology. The neural network is configured such that information input to the input layer is propagated (arithmetic) to the intermediate layer and the output layer in this order, so that an event is output from the output layer. Output the contents.

  For example, the intermediate layer is configured by a plurality of intermediate units Zj (j = 1 to n (the number of units)). Then, the information input to the input units Xi (i = 1 to k (the number of units)) of the input layer is weighted (integrated) by the respective coupling coefficients Wji (not shown), and each intermediate element of the intermediate layer is weighted. The values are input to the unit Zj, and they are added to become the value of each intermediate unit Zj.

  Further, the value of each intermediate unit Zj of the intermediate layer is nonlinearly transformed by an input / output function (for example, a sigmoid function), weighted (integrated) by each coupling coefficient Vj (not shown), and The values are input to the output unit U, and they are added to become the value of the output unit U in the output layer.

  The weighting coefficients between neurons in the neural network are optimized by supervised learning using analysis data in which data input to these input layers and data output from the output layer are linked.

The event contents are calculated by the forward propagation process of the classification model. The configuration of the classification model described above is an example, and may be variously changed. Further, the classification model may further have a pre-processing unit normalization processing of data of the response characteristic applied to the input unit X ₁ to X _n.

<Explanation of regression model>
FIG. 9 shows a regression model.
The regression model is a learning model that outputs the most probable event numerical information from the parameters of the time-series variation of the measured blood glucose data based on the event content estimated by the classification model.

  The parameters of the time-series fluctuation of the blood glucose data include the gradient of the blood glucose fluctuation, the difference between the peak and the blood glucose level at the start of the blood glucose fluctuation, the time until the peak of the blood glucose fluctuation, the difference between the slope of the blood glucose fluctuation until then, There are a blood sugar level at the time point, a blood sugar level at the end of the fluctuation, and the like. The event contents accompanied by the event numerical information include, for example, exercise, meals, insulin administration, and medication.

  The event numerical information to be output includes exercise duration, load, and the like in the case of exercise. For meals, there is a carb count. For insulin administration, there is an insulin dose. And for medication, there is a dosage.

FIG. 10 is a diagram illustrating an example of a configuration of a regression model (here, logistic regression) of the present embodiment.
The regression model according to the present embodiment has, as an input layer, a plurality of input units for inputting parameters of time-series variation of measured blood glucose level data.

The plurality of input units X _{1 to} X _n are respectively used as parameters of the time series variation of the measured blood glucose level data, as a gradient of the blood glucose variation, a difference between the peak and the blood glucose level at the start of the blood glucose variation, and a time until the peak of the blood glucose variation. , The difference in slope from the blood glucose fluctuation, the blood glucose level at the start of the fluctuation, and the blood glucose level at the end of the fluctuation.

  The other configuration of the logistic regression is the same as that of a known technique, and information input to input units Xi (i = 1 to n (the number of units)) is weighted by respective coupling coefficients Wji (not shown). (Integrated), the values are added to a constant, and the result is the value of the output unit U.

The constant to be added to the weighting coefficient is optimized by supervised learning using analysis data in which data input to these input layers and data output from the output layer are linked.
Event numerical information is calculated by regression processing of a regression model. The configuration of the regression model described above is an example, and may be variously changed. Further, the regression model may further include a preprocessing unit for normalizing the response feature data input to the input units Xi to Xk.
For the machine learning of the regression model, a model capable of outputting numerical values by a mathematical expression such as logistic regression shown in FIG. 10 is used.

<Control block diagram and flowchart of event estimation system using classification model>
FIG. 11 shows a simplified control block diagram of an event estimation system including the server device 21 and the terminal device 4.

  Note that FIG. 11 shows a simplified configuration of both the server device 21 and the terminal device 4. The detailed configuration of the server device 21 is shown in FIG. 17 described later, and the detailed configuration of the terminal device 4 is shown in FIG. 18 described later.

  First, in the server device 21, the receiving unit 22 stores first blood glucose level data including blood glucose level data of a plurality of diabetic patients recorded in a database or the like, and event contents related to the date and time of measurement of the first blood glucose level data. For a predetermined period.

  Next, the analysis unit 23 calculates a time-series variation parameter of the first blood glucose level data for the data from the reception unit 22. Then, the analysis unit 23 analyzes the relationship between the first blood sugar level data and the event content by machine learning, and generates a classification model.

  Next, on the terminal device 4 side, the estimating unit (event estimating unit) 24 is the blood glucose level data of the specific user received by the communication unit 31 from the human body-mounted sensor device 1 of the continuous blood glucose measurement (CGM) system. A parameter of a time-series variation of the blood glucose level for a predetermined period of the second blood glucose level data is calculated. Then, the estimating unit 24 uses the classification model to estimate the content of the event related to the measurement date and time of the second blood sugar level data, using the parameter of the time series variation of the blood sugar level as an input.

The classification model is configured to be adjustable in order to further increase the estimation accuracy of the blood glucose level data of a specific user.
In the adjustment unit 25, the second blood sugar level data input by the specific user and received in the communication unit 31 and the event content related to the measurement date and time of the second blood sugar level data are input, and the classification model is adjusted. I do. The adjusted classification model is sent from the adjustment unit 25 to the estimation unit 24.

FIG. 12 shows a flowchart of the event estimation system.
First, in the server device 21, the receiving unit 22 includes first blood glucose level data including blood glucose level data of a plurality of diabetics recorded in a database or the like, and event content related to the date and time of measurement of the first blood glucose level data. Is received for a predetermined period of time (S1).

  Next, the analysis unit 23 calculates a time-series variation parameter of the first blood glucose level data for the data from the reception unit 22. Then, the analysis unit 23 analyzes the relationship between the first blood sugar level data and the event content by machine learning, and performs a second step of generating a classification model.

  Next, the estimating unit 24 calculates the time of the blood glucose level for a predetermined period of the second blood glucose level data which is the blood glucose level data of the specific user received from the human body-mounted sensor device 1 of the continuous blood glucose measurement (CGM) system. Calculate the parameters of the series variation. Then, the estimating unit 24 performs a third step of estimating the event content related to the measurement date and time of the second blood glucose level data using the classification model by using the parameter of the time series variation of the blood glucose level as an input. I do.

<Control block diagram and flowchart of event estimation system using classification model and regression model>
FIG. 13 shows a control block diagram of the event estimation system.

  First, on the server device 21 side, in the receiving unit 22, first blood glucose data including blood glucose data of a plurality of diabetics recorded in a database or the like, and an event related to the measurement date and time of the first blood glucose data. The contents and the event numerical information are received for a predetermined period.

Next, the analyzing unit 23 calculates a parameter of a time-series variation of the first blood sugar level data received from the receiving unit 22.
Then, the analysis unit 23 analyzes the relationship between the first blood sugar level data and the event content by machine learning, and generates a classification model. Then, the relationship between the parameter of the time-series variation of the blood glucose level data and the event numerical information is analyzed by machine learning, and a regression model is generated.

  Next, on the terminal device 4 side, the estimating unit 24 uses the second blood glucose level, which is the blood glucose level data of the specific user, received by the communication unit 31 from the human body-mounted sensor device 1 of the continuous blood glucose measurement (CGM) system. A time-series variation parameter of the blood glucose level for a predetermined period of data is calculated. Then, the estimating unit 24 estimates the event content related to the measurement date and time of the second blood sugar level data by using the classification model as a parameter of the time series variation of the blood sugar level. Further, the estimating unit 24 estimates event numerical value information related to the event content using a regression model.

The classification model is adjusted to improve the estimation accuracy of the blood glucose level data of a specific user.
In the adjustment unit 25, the second blood sugar level data input by the specific user, the event content related to the measurement date and time of the second blood sugar level data, and the event numerical information related to the event content are input. And adjust the classification and regression models. Then, the classification model and the regression model adjusted by the adjustment unit 25 are sent from the adjustment unit 25 to the estimation unit 24.

FIG. 14 shows a flowchart of the event estimation system.
First, on the server device 21 side, the receiving unit 22 stores first blood sugar level data including blood sugar level data of a plurality of diabetic patients recorded in a database or the like, and an event related to the measurement date and time of the first blood sugar level data. A first step of receiving the contents and the event numerical information for a predetermined period is performed (S4).

  Next, the analysis unit 23 calculates a time-series variation parameter of the first blood glucose level data for the data from the reception unit 22. Then, the analysis unit 23 analyzes the relationship between the first blood sugar level data and the event content by machine learning, generates a classification model, and relates the relationship between the time series variation parameter of the blood sugar level data and the event numerical information. Is analyzed by machine learning, and a second step of generating a regression model is performed (S5).

  Next, on the side of the terminal device 4, the estimation unit 24 determines the predetermined period of the second blood glucose level data, which is the blood glucose level data of the specific user, received from the human body-mounted sensor device 1 of the continuous blood glucose measurement (CGM) system. The parameter of the time series fluctuation of the blood glucose level of the minute is calculated. Then, the estimating unit 24 uses the classification model as an input, estimates the event content related to the measurement date and time of the second blood glucose level data, using the time series variation parameter of the blood glucose level as an input, and uses the regression model. Then, a third step of estimating event numerical information related to the event content is performed (S6).

<Flowchart of analysis unit>
FIG. 15 shows a flowchart of the generation process of the classification model and the regression model in the analysis unit 23 on the server device 21 side.

  The flowchart of FIG. 15 uses a system configuration using a human-worn sensor device 1, a terminal device 4, and a server device 21 of a continuous blood glucose measurement (CGM) system. It is an example of a flow of generating a classification model and a regression model in the analysis unit 23 provided in the server device 21 in this configuration.

First, the terminal device 4 causes the obtaining unit 33 to obtain blood glucose level data from the human body-mounted sensor device 1 of the continuous blood glucose measurement (CGM) system (S7).
Next, via the input unit 34 of the terminal device 4, the user inputs the event content and the event numerical information at the time of measuring the blood sugar level (S8).

Next, the terminal device 4 transmits the blood glucose level data, the event content at the time of blood glucose level measurement, and the event numerical value information to the server device 21 via the communication unit 31 (S9).
The terminal device 4 transmits the blood sugar level data, the event content at the time of blood sugar level measurement, and the event numerical value information to the server device 21 via the communication unit 31 for a predetermined period set in advance (S10).

  Next, in the server device 21, the analysis unit 23 performs machine learning using the received blood glucose level data for a predetermined period, the event content and the event numerical information at the time of blood glucose level measurement, and generates a classification model and a regression model. (S11).

Next, the server device 21 transmits the classification model and the regression model to the terminal device 4 via the communication unit 26 (S12).
Next, in the terminal device 4, the classification model and the regression model received via the communication unit 31 are set in the estimation unit 24 (S13).

<Flowchart of adjustment unit>
FIG. 16 shows a flowchart of the adjustment processing in the adjustment unit 25 of the terminal device 4.
First, in the terminal device 4, the acquiring unit 33 acquires the second blood glucose level data from the human body-mounted sensor device 1 of the continuous blood glucose measurement (CGM) system (S14).

Next, in the terminal device 4, the estimating unit 24 estimates event contents and event numerical information using the set classification model and regression model (S15).
Next, the adjusting unit 25 determines whether or not the estimated event content and event numerical information need to be corrected (S16). This determination may be made by the user to make a determination, or the terminal device 4 itself may determine whether correction is necessary based on the estimated contents and the user's determination.

Next, when correction is necessary, the user inputs the event content and the event numerical information via the input unit 34 of the terminal device 4 (S17).
Next, the adjustment unit 25 of the terminal device 4 adjusts the classification model and the regression model based on the input event contents and event numerical information (S18).

  Specifically, by supervised learning using data input to these input layers and data for analysis linked to data output from the output layer, weighting coefficients between neurons of the neural network, The constant to be added to the weighting coefficient of the logistic regression is again optimized and adjusted.

Next, the estimation unit 24 updates the already set classification / regression model to the adjusted classification model and regression model (S19).
As described above, the adjustment unit 25 performs the transfer learning on the classification model and the regression model as the learned models with respect to the blood glucose level data of the specific user.

<Description of server device of event estimation system>
FIG. 17 is a control block diagram of the server device 21 included in the event estimation system according to the present embodiment.

As shown in FIG. 17, the server device 21 includes an analysis unit 23, a communication unit 26, a control unit 27, a calculation unit 28, an output unit 29, and a storage unit 30.
As described with reference to FIGS. 11 and 13, the analysis unit 23 analyzes the relationship between the first blood glucose level data and the event content by machine learning, generates a classification model, and generates a classification model. The relationship with the event numerical information is analyzed by machine learning to generate a regression model.

The communication unit 26 includes the receiving unit 22 shown in FIGS. 11 and 13 and transmits and receives data and the like to and from the terminal device 4 (communication unit 31).
Then, the communication unit 26 (the receiving unit 22) compares the first blood sugar level data, the event content related to the measurement date and time of the first blood sugar level data, and the event numerical information related to the event content for a predetermined period of time. Receive.

  The control unit 27 controls the first blood sugar level data received by the communication unit 26 (the receiving unit 22), the event content (type) related to the measurement date and time of the first blood sugar level data, and the event numerical value related to the event content. The information is recorded in the storage unit 30 for a predetermined period. Then, the control unit 27 inputs the recorded data to the calculation unit 28.

Further, the control unit 27 outputs the classification model and the regression model from the output unit 29 to the network via the communication unit 26.
The calculation unit 28 calculates a parameter of a time-series variation of the first blood sugar level data.

The output unit 29 transmits the classification model and the regression model received from the control unit 27 and the communication unit 26.
As described above, the storage unit 30 stores the first blood sugar level data for a predetermined period, the event content, the event numerical information, and the like.

<Description of terminal device (user terminal device) of event estimation system>
FIG. 18 shows a control block diagram of the terminal device 4 included in the event estimation system of the present embodiment.

  As illustrated in FIG. 18, the terminal device 4 as a user terminal device (for example, a smartphone or the like) includes an estimation unit 24, an adjustment unit 25, a communication unit 31, a control unit 32, an acquisition unit 33, and an input unit. 34, a notification unit 35, a display unit 36, and an audio output unit 37.

  As described in FIGS. 11 and 13, the estimating unit 24 is the second blood glucose which is the blood glucose level data of the specific user received by the communication unit 31 from the human body-mounted sensor device 1 of the continuous blood glucose measurement (CGM) system. A parameter of a time-series variation of the blood glucose level for a predetermined period of the value data is calculated. Then, the estimating unit 24 uses the classification model to estimate the content of the event related to the measurement date and time of the second blood sugar level data, using the parameter of the time series variation of the blood sugar level as an input.

  As described with reference to FIGS. 11 and 13, the adjustment unit 25 also receives the second blood glucose level data input by the specific user and received by the communication unit 31 and the event content related to the measurement date and time of the second blood glucose level data. Is input to adjust the classification model. Then, the adjustment unit 25 transmits the adjusted classification model to the estimation unit 24.

  The communication unit 31 communicates with the human-worn sensor device 1 via short-range communication means (for example, BLE (Bluetooth (registered trademark) Low Energy)), and also communicates with the server device 21 via a network such as the Internet. Perform communication.

  The control unit 32 sets the classification model and the regression model received from the server device 21 via the communication unit 31 in the estimation unit 24. Further, the control unit 32 receives the second blood glucose level data from the human body-mounted sensor device 1 via the communication unit 31 and transmits the data to the acquisition unit 33. Further, the control unit 32 causes the estimating unit 24 to estimate the content of the event using the set classification model based on the second blood sugar level data transmitted to the acquiring unit 33, and to use the regression model to generate the event numerical information. Is estimated.

The user inputs the event content and event numerical information related to the event content to the input unit 34.
The notifying unit 35 notifies (requests) the user to input an event when the estimating unit 24 estimates that the event has occurred.

  The display unit 36 is, for example, a device that displays various kinds of information by a liquid crystal display panel. Is displayed.

The voice output unit 37 is a device that outputs voice or sound, and outputs, for example, a message that prompts the user to input an event by voice.
When the estimated event content and event numerical value information need to be corrected, the terminal device 4 of the present embodiment and the second blood sugar level data and the second blood sugar level input from the input unit 34 by the user. The event data related to the measurement date and time of the value data is applied to the classification model, and the adjustment unit 25 adjusts the classification model. Further, the terminal device 4 applies the event content input from the input unit 34 and the event numerical information related to the event content to the regression model, and the adjustment unit 25 adjusts the regression model.

  As described above, the event estimation system according to the present embodiment includes the server device 21, the terminal device 4, and the network that connects the server device 21 and the terminal device 4 so that they can communicate with each other.

<Event input notification function of terminal device>
Here, the event input notification function in the terminal device 4 will be described again using the control block diagram of the terminal device 4 of the event estimation system in FIG.

As described above, the terminal device 4 of the present embodiment estimates an event based on the second blood sugar level data received from the human body-mounted sensor device 1.
Further, as described above, the terminal device 4 according to the present embodiment further includes the notification unit 35 that notifies (requests) the user to input an event when it is estimated that the event has occurred. .

Further, the terminal device 4 causes the display unit 36 to display the blood glucose level data of the user as the biological information.
As a method of notification by the notification unit 35, a method of displaying a message urging the user to input an event on the display unit 36, or a method of outputting a voice or sound from the audio output unit 37 to notify the user is conceivable. As another method, the user may be notified by vibrating the terminal device 4 itself.

As a method of notification, the sound, the display, and the vibration may be appropriately combined for notification.
Furthermore, when notifying the user, the notifying unit 35 may display the event estimated in advance on the display unit 36 to prompt the user to confirm. In this case, the control unit 32 transmits the event estimated by the estimation unit 24 to the notification unit 35. Then, the notification unit 35 prompts the user to input an event while displaying the estimated event.

If the user does not input an event even after the notification, the notification unit 35 can repeatedly notify the user.
That is, when the user does not input an event from the input unit 34 in spite of the event input request from the notification unit 35, the control unit 32 transmits the event from the notification unit 35 to the user again. Then, an event input from the input unit 34 is requested.

Here, in the event estimation system of the present embodiment, a configuration may be adopted in which an external device communicably provided via the communication unit 31 notifies an input of an event.
The external device is preferably in a user-friendly form. For example, a robot type (mascot type) modeled on a child or a pet may be used to notify the occurrence of an event using voice or a display screen and to prompt an event input.

  Specifically, FIG. 26 shows an example of a robot-type external device 39 modeled on a child or pet.

  In the external device 39 shown in FIG. 26, in response to a voice output of “Did you eat rice?”, When a human reply “Yes, I understand well.”, “Delicious? What did you eat?” During the conversation, the user is notified of the occurrence of an event or prompted to input an event.

  Thereby, the user such as a diabetic patient can provide the necessary information to the external device 39 in a natural conversation with the robot-type external device 39 and is prompted to input a necessary event. Can be done.

  The robot-type external device 39 is not limited to requesting an event input by voice output, and may be configured to request an event input as character information on a display unit (not shown).

<Example of event input notification function of terminal device>
FIGS. 19A to 19D show an example of a display screen showing the event input notification function of the terminal device 4.

In the examples shown in FIGS. 19A to 19D, the estimating unit 24 estimates the occurrence of the exercise event in the measured blood sugar level data.
The display screen shown in FIG. 19A shows a state in which the occurrence of an exercise event is estimated and displayed on the display unit 36 to prompt the user to input the event. If the user has exercised as estimated, the user touches OK.

The display screen shown in FIG. 19B shows a state in which the user is prompted to input exercise intensity and exercise time as event numerical information of the exercise event.
The display screen shown in FIG. 19C shows a state where the user is inputting the event numerical information of the exercise event.

  The display screen shown in FIG. 19D shows an event input completion screen. In the fifth split screen 14e, an exercise event, exercise intensity, and time of exercise are input as the latest events.

  Next, in the examples shown in FIGS. 20A and 20B, the estimation unit 24 displays the measured blood sugar level data on the terminal device 4 that estimates the occurrence of the drug treatment (medication) event. Shows the screen.

  The display screen shown in FIG. 20A shows a state in which the occurrence of a drug treatment (medication) event is estimated and displayed on the display unit 36 to prompt the user to input the event. The user touches OK if the drug treatment (medication) has been performed as estimated.

  The display screen illustrated in FIG. 20B is a display screen that prompts the user to input a drug name (insulin) and its dose (unit (s)) as event numerical information in a drug treatment (medication) event.

  In the terminal device 4 of the present embodiment, by displaying the display screen as described above, the control unit 32 prompts the user to input an event from the notification unit 35 when the estimation unit 24 estimates that the event has occurred. Request that This prevents the user from forgetting to input the event, and also allows the user to input the event with only a simple input operation, thereby improving the convenience of the biological information measuring device. be able to.

<Flowchart of event input notification function of terminal device>
FIG. 21 is a flowchart illustrating a flow of the event input notification process of the terminal device 4.
In the terminal device 4 of the present embodiment, the acquisition unit 33 acquires the blood glucose level data from the human body-worn sensor device 1 of the continuous blood glucose measurement (CGM) system via the communication unit 31 (S20).

Next, in the terminal device 4, the estimation unit 24 calculates a parameter of a time-series variation of the second blood sugar level data (S21).
Next, the terminal device 4 estimates the event content (type) and the event numerical information using the classification model and the regression model set in the estimation unit 24 (S22).

Next, when it is estimated that an event has occurred (S23), the control unit 32 sends a notification (request) from the notification unit 35 to prompt the user to input the event (S24).
As a notification method of the notification unit 35, a message urging the user to input an event may be displayed on the display unit 36. As another notification method, a sound or a sound may be output from the sound output unit 37 to notify the user. As still another method, the user may be notified by vibrating the terminal device 4 itself (S25).
Finally, the control unit 32 receives an input to the input unit 34 from the user who has been prompted to input the event (S26).

<Automatic input function for terminal device events>
FIG. 22 shows a control block diagram of the terminal device 4 having an automatic event input function.
As described above, the terminal device 4 of the present embodiment estimates an event based on the second blood sugar level data received from the human body-mounted sensor device 1.
Then, here, when the estimation unit 24 estimates that the event has occurred, the terminal device 4 automatically inputs the event without the intervention of the user.

The terminal device 4 of the present embodiment further includes a confirmation notification unit 38 that requests confirmation of the content of the event when the event is automatically input.
The terminal device 4 includes a display unit 36 for displaying blood glucose level data of the user as biological information.

  As a method of notification by the confirmation notification unit 38, a method of displaying a message urging the user to input an event on the display unit 36, or a method of outputting a voice or sound from the audio output unit 37 to notify the user is conceivable. . As another method, the user may be notified by vibrating the terminal device 4 itself.

As a method of notification, the sound, the display, and the vibration may be appropriately combined for notification.
Further, when notifying the user, the confirmation notification unit 38 estimates an event in advance, causes the display unit 36 to display the event automatically input by the input unit 34, and displays the event automatically input by the user. You may be prompted to confirm whether or not it is appropriate. In this case, the control unit 32 transmits the content (type) of the event estimated by the estimation unit 24 to the confirmation notification unit 38. Then, the confirmation notification unit 38 prompts the user to confirm whether or not the input event is appropriate while displaying the event automatically input by the input unit 34.

If the user does not check the automatically input event even after the notification, the confirmation notifying unit 38 can also repeatedly notify the user of the confirmation.
That is, when the user does not perform the operation of confirming the content of the event in spite of the confirmation request from the confirmation notifying unit 38, the control unit 32 sends the event to the user again from the confirmation notifying unit 38. Request confirmation of the contents.

Here, in the event estimation system of the present embodiment, a confirmation of an automatically input event is notified in an external device communicably provided via the communication unit 31.
The external device is preferably in a user-friendly form. For example, a robot type (mascot type) modeled on a child or pet, a device that notifies the contents of an automatically input event using voice or a display screen, and prompts confirmation of the contents of the automatically input event. It may be.
Specifically, as described above, a robot-type external device 39 modeled on a child or pet may be used (see FIG. 26).

<Example of automatic input notification function of event of terminal device>
FIGS. 23A to 23C show an example of a display screen showing an event automatic input notification function of the terminal device 4.
In the examples shown in FIGS. 23A to 23C, the estimation unit 24 estimates the occurrence of the exercise event in the measured blood glucose level data.
In the display screen shown in FIG. 23A, after the estimating unit 24 estimates the occurrence of the exercise event, the control unit 32 causes the input unit 34 to automatically input the estimated event, and displays a confirmation screen on the display unit 36. This prompts the user to confirm the event. The user checks this display screen, and touches OK if the user has exercised as estimated.

The display screen shown in FIG. 23B shows that the intensity (light) of the exercise and the time (30 minutes) of the exercise are automatically input as the numerical value information of the exercise event.
On the display screen shown in FIG. 23C, the event numerical information of the exercise event is automatically input. In the fifth divided screen 14e, the exercise event, the intensity of the exercise, and the time of the exercise are automatically input as the latest events.

  Next, in the example shown in FIGS. 24A and 24B, the estimation unit 24 estimates the occurrence of a drug treatment (medication) event based on the measured blood glucose level data, The input display screen is shown.

  In the display screen shown in FIG. 24A, the occurrence of the estimated drug treatment (medication) event is automatically input, and a confirmation screen thereof is displayed on the display unit 36 to urge the user to confirm the event. . The user touches OK if the drug treatment (medication) has been performed as estimated.

  The display screen illustrated in FIG. 24B is a display screen that prompts confirmation of a drug name (insulin) and a dose (unit (s)) as event numerical information in a drug treatment (medication) event.

  In the terminal device 4 of the present embodiment, the control unit 32 automatically displays the estimated event by the input unit 34 when the estimation unit 24 estimates the occurrence of the event by displaying the display screen as described above. Then, the user is requested to confirm the event from the confirmation notification unit 38. Thus, it is possible to prevent the user from forgetting to input the event, and the user only needs to check the content of the automatically input event, thereby improving the convenience of the biological information measuring device. be able to.

<Flowchart of event input notification function of terminal device>
FIG. 25 is a flowchart illustrating a flow of an event automatic input notification process of the terminal device 4.
In the terminal device 4 of the present embodiment, the acquisition unit 33 acquires the blood glucose level data from the human body-worn sensor device 1 of the continuous blood glucose measurement (CGM) system via the communication unit 31 (S27).
Next, in the terminal device 4, the estimation unit 24 calculates a parameter of a time-series variation of the second blood sugar level data (S28).

Next, the terminal device 4 estimates the event content and the event numerical value information using the classification model and the regression model set in the estimation unit 24 (S29).
Next, when it is estimated that an event has occurred (S30), the control unit 32 causes the estimating unit 24 to estimate the estimated event numerical information (S31).

  Then, the event estimated by the estimating unit 24 is automatically input by the input unit 34 (S32). Then, the notification unit 35 notifies the user of the event occurrence and the event content automatically received by the input unit 34 and automatically input by the input unit 34, and requests confirmation of the content of the automatically input event (S33).

  As a method of notification by the confirmation notification unit 38, a message urging the user to confirm the content of the event automatically input may be displayed on the display unit 36. As another notification method, a voice or a sound may be output from the voice output unit 37 to notify the user of a notification prompting confirmation of the content of the automatically input event. As another method, the user may be notified by vibrating the terminal device 4 itself (S25).

Finally, the control unit 32 accepts a user's input via the input unit 34.
Further, in the terminal device 4 of the present embodiment, the input unit 34 accepts a confirmation operation by the user who has confirmed the content of the event automatically input, and the control unit 32 causes the control unit 32 to The display may be changed according to the operation.

More specifically, when the determination operation is received, the control unit 32 displays “determination” so that the content of the event for which the determination operation has been received can be easily recognized as having been determined. Good.
[Other embodiments]
As described above, one embodiment of the present invention has been described, but the present invention is not limited to the above embodiment, and various changes can be made without departing from the gist of the invention.

(A)
In the above embodiment, an example has been described in which the present invention is implemented as an event estimation method, a server device, a biological information measurement device, and an event estimation system. However, the present invention is not limited to this.

For example, the present invention may be realized as an event estimation program that causes a computer to execute the event estimation method described in the above embodiment.
The event estimation program may be stored in a storage unit of a server device or the like, and is read by hardware such as a CPU, so that the computer can execute the above-described event estimation method.

  The event estimation method of the present invention is greatly expected to be applied to, for example, event estimation relating to measurement of a blood glucose level by a blood glucose level sensor included in a continuous blood glucose measurement system.

DESCRIPTION OF SYMBOLS 1 Human body-mounted sensor device 2 Upper arm part 3 Electronic unit 4 Terminal device (biological information measuring device)
Reference Signs List 5 Sensor 6 Mount unit 7 Sensor unit 8 Measurement unit 9 Control unit 10 Temperature sensor 11 Storage unit 12 Communication unit 13 Battery 14a First split screen 14b Second split screen 14c Third split screen 14d Fourth split screen 14e Fifth split screen 14f Sixth divided screen 14g Seventh divided screen 20 Database 21 Server device 22 Receiving unit 23 Analyzing unit 24 Estimating unit (event estimating unit)
25 adjustment unit 26 communication unit 27 control unit 28 calculation unit 29 output unit 30 storage unit 31 communication unit 32 control unit 33 acquisition unit 34 input unit 35 notification unit 36 display unit 37 audio output unit 38 confirmation notification unit 39 external device U output unit Vj, Wji Coupling coefficient Xi Input unit Zj Intermediate unit

Claims (30)

A first step of receiving, for a predetermined period, first blood sugar level data and event contents related to the measurement date and time of the first blood sugar level data;
A second step of calculating a parameter of a time-series variation of the first blood sugar level data, analyzing a relationship between the first blood sugar level data and the event content by machine learning, and generating a classification model;
Using the parameters of the time series variation of the blood glucose level for a predetermined period of the received second blood glucose level data as input, and using the classification model, the event content related to the measurement date and time of the second blood glucose level data A third step of estimating;
Event estimation method with A first step of receiving first blood sugar level data, event content related to the measurement date and time of the first blood sugar level data, and event numerical value information for a predetermined period;
Calculating a parameter of a time-series variation of the first blood sugar level data, analyzing a relationship between the first blood sugar level data and the content of the event by machine learning, generating a classification model, and generating a classification model; A second step of analyzing, by machine learning, the relationship between the parameter of the time-series variation and the event numerical information, and generating a regression model;
Using the classification model as an input, input the parameters of the time-series variation of the blood glucose level for a predetermined period of the received second blood glucose level data, and estimate the event content related to the measurement date and time of the second blood glucose level data. And a third step of using the regression model to estimate event numerical information related to the event content;
Event estimation method with The parameters of the time-series fluctuation of the blood glucose level are the gradient of the blood glucose fluctuation, the difference between the peak and the blood glucose level at the start of the blood glucose fluctuation, the time until the peak of the blood glucose fluctuation, the difference between the slope with the blood glucose fluctuation until then, and the time when the fluctuation starts. Including at least one of a blood glucose level and a blood glucose level at the end of fluctuation.
An event estimation method by monitoring a blood sugar level according to claim 1 or 2. The event content is selected from an event set including at least one of hypoglycemia awareness, sick day, menstrual day, wake-up, exercise, bathing, meal, bedtime, insulin administration, medication.
The event estimation method according to claim 1. The first blood sugar level data is blood sugar level data of a plurality of persons,
The event estimation method according to claim 1. The second blood sugar level data is blood sugar level data of a specific diabetic patient,
The event estimation method according to claim 1. Receiving the second blood sugar level data and the event content related to the date and time of measurement of the second blood sugar level data, applying the event to the generated classification model, and corresponding to the second blood sugar level data; Further comprising a step of adjusting the classification model,
The event estimation method according to claim 1. The event numerical information is information related to the estimated event,
The event estimation method according to claim 2. Receiving the second blood sugar level data, the event content related to the measurement date and time of the second blood sugar level data, and the event numerical information related to the event content; and generating the classification model and the regression. And a step of adjusting the regression model and the classification model corresponding to the second blood sugar level data, the method further comprising:
The event estimation method according to any one of claims 2 to 6, and 8. A first step of receiving, for a predetermined period, first blood sugar level data and event contents related to the measurement date and time of the first blood sugar level data;
A second step of calculating a parameter of a time-series variation of the first blood sugar level data, analyzing a relationship between the first blood sugar level data and the event content by machine learning, and generating a classification model;
Using the classification model as an input, input the parameters of the time-series variation of the blood glucose level for a predetermined period of the received second blood glucose level data, and estimate the event content related to the measurement date and time of the second blood glucose level data. A third step,
An event estimation program for causing a computer to execute an event estimation method including: A first step of receiving first blood sugar level data, event content related to the measurement date and time of the first blood sugar level data, and event numerical value information for a predetermined period;
Calculating a parameter of a time-series variation of the first blood sugar level data, analyzing a relationship between the first blood sugar level data and the content of the event by machine learning, generating a classification model, and generating a classification model; A second step of analyzing, by machine learning, the relationship between the parameter of the time-series variation and the event numerical information, and generating a regression model;
Using the classification model as an input, input the parameters of the time-series variation of the blood glucose level for a predetermined period of the received second blood glucose level data, and estimate the event content related to the measurement date and time of the second blood glucose level data. And a third step of using the regression model to estimate event numerical information related to the event content;
An event estimation program for causing a computer to execute an event estimation method including: The parameters of the time-series fluctuation of the blood glucose level are the gradient of the blood glucose fluctuation, the difference between the peak and the blood glucose level at the start of the blood glucose fluctuation, the time until the peak of the blood glucose fluctuation, the difference between the slope with the blood glucose fluctuation until then, and the time when the fluctuation starts. Including at least one of a blood glucose level and a blood glucose level at the end of fluctuation.
The event estimation program according to claim 10. The event content is selected from an event set including at least one of hypoglycemia awareness, sick day, menstrual day, wake-up, exercise, bathing, meal, bedtime, insulin administration, medication.
The event estimation program according to any one of claims 10 to 12. The first blood sugar level data is blood sugar level data of a plurality of persons,
The event estimation program according to any one of claims 10 to 13. The second blood sugar level data is blood sugar level data of a specific diabetic patient,
The event estimation program according to any one of claims 10 to 14. The second blood sugar level data and the event content related to the measurement date and time of the second blood sugar level data are received and applied to the generated classification model, and correspond to the second blood sugar level data. Further comprising a step of adjusting the classification model,
The event estimation program according to any one of claims 10, 12 to 15. The event numerical information is information related to the estimated event,
The event estimation program according to any one of claims 11 to 15. Receiving the second blood sugar level data, the event content related to the measurement date and time of the second blood sugar level data, and the event numerical information related to the event content; and generating the classification model and the regression. And adjusting the regression model and the classification model corresponding to the second blood sugar level data, the method further comprising:
The event estimation program according to any one of claims 11 to 15, 17. A receiving unit that receives first blood sugar level data and event contents related to the measurement date and time of the first blood sugar level data for a predetermined period;
An analysis unit that calculates a parameter of a time-series variation of the first blood sugar level data, analyzes a relationship between the first blood sugar level data and the event content by machine learning, and generates a classification model;
An output unit that outputs the classification model generated by the analysis unit,
Server device equipped with A receiving unit that receives first blood sugar level data, event content related to the measurement date and time of the first blood sugar level data, and event numerical information related to the event content for a predetermined period;
Calculating a parameter of a time-series variation of the first blood sugar level data; analyzing a relationship between the first blood sugar level data and the event content by machine learning; generating a classification model; An analysis unit that analyzes the relationship between the event numerical information related to the event content by machine learning and generates a regression model,
An output unit that outputs the classification model and the regression model generated by the analysis unit,
Server device equipped with The parameters of the time-series fluctuation of the blood glucose level are the gradient of the blood glucose fluctuation, the difference between the peak and the blood glucose level at the start of the blood glucose fluctuation, the time until the peak of the blood glucose fluctuation, the difference between the slope with the blood glucose fluctuation until then, and the time when the fluctuation starts. Including at least one of a blood glucose level and a blood glucose level at the end of fluctuation.
The server device according to claim 19. The event content is selected from an event set including at least one of hypoglycemia awareness, sick day, menstrual day, wake-up, exercise, bathing, meal, bedtime, insulin administration, medication.
The server device according to claim 19. The first blood sugar level data is blood sugar level data of a plurality of persons,
The server device according to claim 19. The event numerical information is information related to the estimated event,
The server device according to claim 20. An acquisition unit that captures second blood glucose level data from the sensor;
An event estimator for estimating the content of the event using a classification model;
A control unit that inputs the second blood sugar level data taken into the acquisition unit to the event estimation unit, and estimates event information related to the second blood sugar level data;
A biological information measuring device comprising: An acquisition unit that captures second blood glucose level data from the sensor;
An event estimator for estimating the content of the event using a classification model and estimating event numerical information using a regression model;
A control unit that inputs the second blood sugar level data taken into the acquisition unit to the event estimation unit, and estimates the content of the event and the event numerical value information related to the second blood sugar level data;
A biological information measuring device comprising: An input section for inputting event content,
An adjustment unit configured to apply the second blood sugar level data and the event content related to the measurement date and time of the second blood sugar level data input from the input unit to the classification model, and adjust the classification model When,
Further equipped with
The biological information measuring device according to claim 25. The second blood sugar level data is blood sugar level data of a specific diabetic patient,
The biological information measuring device according to claim 25 or 26. An input unit for inputting event content and event numerical information related to the event content;
Applying the second blood sugar level data and the event content related to the measurement date and time of the second blood sugar level data input from the input unit to the classification model, and adjusting the classification model, The event content input from the input unit, and the event numerical information related to the event content, applied to the regression model, further comprising an adjustment unit for adjusting the regression model,
The biological information measuring device according to claim 26. A server device according to any one of claims 19 to 24,
The biological information measuring device according to any one of claims 25 to 29,
A network that communicably connects the server device and the biological information measurement device,
Event estimation system with.

Images