KR20210015045A - Glucose measuring apparatus and measuring system using the same - Google Patents

Abstract

A blood glucose measurement device includes: a measuring unit which acquires a current signal obtained by applying a voltage to an electrode which is formed with a porous platinum layer and is provided minimally invasively on skin; a current characteristic search unit which extracts characteristic values from the acquired current signal; and a blood glucose measurement unit which measures a blood glucose value from the feature values extracted by the current characteristic search unit through a generated machine learning model. Therefore, continuous blood glucose measurement is possible.

Description

Blood glucose measuring apparatus and blood glucose measuring system using the same

The present invention relates to a blood glucose measurement apparatus and a blood glucose measurement system using the same.

Diabetes mellitus is a symptom in which blood sugar stays at a higher level than normal, and it is mainly caused by not secreting insulin or not using insulin in cells. According to clinical studies, it is reported that the death from diabetes decreased by 12% and various complications by 25-76% when the active blood sugar management therapy was performed through frequent blood sugar measurement.

In addition, in diabetic patients, there is a risk of sudden hypoglycemic shock due to drug overuse. Therefore, it can be said that it is very important to keep the blood sugar within the normal range through continuous blood sugar measurement.

There are two types of blood glucose measurement: a single-use measuring device and a continuous measuring device. When blood sugar is measured with a disposable strip sensor, which is one of the disposable measuring devices, even if it is temporarily high blood sugar or low blood sugar, there is a case where it is not recognized and excessive. In contrast, a continuous blood glucose measurement device can provide such risk information to patients and doctors in real time, enabling more effective blood glucose management.

An aspect of the present invention provides a blood glucose measurement apparatus capable of continuously measuring blood glucose and a blood glucose measurement system using the same.

An aspect of the present invention provides a blood glucose measurement device that minimizes invasion and a blood glucose measurement system using the same.

An aspect of the present invention provides a blood glucose measurement apparatus with improved accuracy and a blood glucose measurement system using the same.

A blood glucose measurement apparatus according to the present invention includes: a measurement unit for obtaining a current signal obtained by applying a voltage to an electrode of a minimally invasive form to the skin having a porous platinum layer formed on the surface thereof; A current characteristic search unit for extracting characteristic values from the obtained current signal; And a blood glucose measurement unit that measures a blood glucose value from the feature values extracted by the current characteristic search unit through the generated machine learning model.

The current characteristic search unit may include a first characteristic extraction unit for extracting a first characteristic, which is a current value at a measurement point, from the obtained current signal; And a second feature extractor for extracting a second feature, which is a correction current value corresponding to the measurement point, from the correction curve generated based on the obtained current signal.

The correction curve may satisfy any one of the following equations.

R= -C ₁ /t+C ₂

R=C ₃ e^(-t)+C ₄

R=-C ₅ log(t)+C ₆

The current characteristic search unit may include a third feature extraction unit for extracting a third characteristic at the time of measurement; It may further include a fourth feature extraction unit for extracting a fourth feature that is the total amount of electric charge up to the measurement point.

The machine learning model is learned from a database in which first data on the acquired current signals for a plurality of subjects and second data on actual blood glucose values of the plurality of subjects are accumulated, and the first, 2 The relationship of data can be learned.

The porous platinum layer may be configured to be formed on the electrode surface.

The electrode may include at least one needle electrode having the porous platinum layer formed on a surface thereof and formed in a needle shape.

The blood glucose measurement system according to the idea of the present invention includes a measuring unit that obtains a current signal obtained by applying a voltage to an electrode provided with a porous platinum layer in a form that is minimally invasive on the skin, and a measurement unit that obtains characteristic values from the obtained current signal A blood glucose measurement device including a current characteristic search unit to extract, and a blood glucose measurement unit measuring a blood glucose value from the feature values extracted by the current characteristic search unit through the generated machine learning model; And a machine learning device that generates the machine learning model.

The blood glucose measurement method according to the idea of the present invention is to obtain a current signal by invading at least one needle electrode with a porous platinum layer on the skin and applying a voltage, extracting feature values from the obtained current signal, and generating a machine Using the learning model, the blood glucose value is measured from the extracted feature values.

The feature values include: a first feature that is a current value at the time of measurement; A second characteristic of a correction current value corresponding to the measurement point in the correction curve generated based on the obtained current signal; A third feature that is the measurement point; It may include a fourth characteristic, which is the total amount of charge up to the measurement point.

According to an aspect of the present invention, it is possible to continuously measure blood glucose.

According to an aspect of the present invention, continuous blood glucose measurement with minimal invasion is possible.

According to an aspect of the present invention, continuous blood glucose measurement with improved accuracy is possible.

According to an aspect of the present invention, since the sterilization operation is easily performed without using an enzyme, hygienic continuous blood glucose measurement is possible.

1 (a) and (b) are diagrams showing a platinum layer of a blood glucose measurement device according to an embodiment of the present invention
2 is a block diagram of a blood glucose measurement apparatus according to an embodiment of the present invention.
3 is a view showing a characteristic value in a measured current graph of a blood glucose measurement device according to an embodiment of the present invention.
Figure 4 is a flow chart of a blood glucose measurement device according to an embodiment of the present invention.
5 is a block diagram of a machine learning apparatus for a blood glucose measurement system according to an embodiment of the present invention.
6 is a diagram for correcting a time delay of a machine learning apparatus of a blood glucose measurement system according to an embodiment of the present invention.
7 is a diagram illustrating a machine learning model generation of a machine learning device of a blood glucose measurement system according to an embodiment of the present invention.
8(a), 8(b), and 8(c) are diagrams showing the accuracy of a machine learning model of a machine learning device of a blood glucose measurement system according to an embodiment of the present invention.

The embodiments described in the present specification and the configurations shown in the drawings are only preferred examples of the disclosed invention, and there may be various modifications that may replace the embodiments and drawings of the present specification at the time of filing of the present application.

In addition, the same reference numerals or reference numerals shown in each drawing of the present specification indicate parts or components that substantially perform the same function.

In addition, terms used in the present specification are used to describe the embodiments, and are not intended to limit and/or limit the disclosed invention. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "comprise" or "have" are intended to designate the presence of features, numbers, steps, actions, components, parts, or a combination thereof described in the specification, but one or more other features. It does not preclude the presence or addition of elements, numbers, steps, actions, components, parts, or combinations thereof.

In addition, terms including ordinal numbers such as “first” and “second” used in the present specification may be used to describe various elements, but the elements are not limited by the terms, and the terms are It is used only for the purpose of distinguishing one component from other components. For example, without departing from the scope of the present invention, a first element may be referred to as a second element, and similarly, a second element may be referred to as a first element. The term “and/or” includes a combination of a plurality of related listed items or any of a plurality of related listed items.

In addition, terms such as "~ unit", "~ group", "~ block", "~ member", and "~ module" may mean a unit that processes at least one function or operation. For example, the terms may refer to at least one hardware such as field-programmable gate array (FPGA)/application specific integrated circuit (ASIC), at least one software stored in a memory, or at least one process processed by a processor. have.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

1(a) and (b) are diagrams showing a platinum layer of a blood glucose measurement apparatus according to an embodiment of the present invention, and FIG. 2 is a block diagram of a blood glucose measurement apparatus according to an embodiment of the present invention.

The blood glucose measurement system 1 is provided to continuously measure the blood glucose of a subject. In the present invention, a blood glucose measurement method that does not use an enzyme may be applied to a blood glucose measurement method. The enzyme-based blood glucose measurement method using enzymes is, in principle, sensitive after implantation in the body as the enzyme loses its activity when the measuring device undergoes high-temperature sterilization, chemical sterilization, or ultraviolet sterilization when an enzyme is used in the measuring device. I can't. Ineffective news The blood glucose measurement method does not use an enzyme, so the sterilization process that is performed before the in vivo insertion process does not affect the performance of the sensor.

Ineffective news The continuous blood glucose measurement method has the advantage of being able to show an accuracy independent of oxygen concentration. In the case of blood glucose collection method, there is a sufficient amount of oxygen in the air, so it is less affected by the oxygen concentration, but the enzymatic continuous blood glucose measurement method has to be inserted into the body with minimal invasiveness, so the concentration of oxygen is lowered and the accuracy is lowered. .

However, the invalid news continuous blood glucose measurement method uses a method of directly oxidizing glucose, so that reliable blood glucose measurement can be performed regardless of oxygen concentration.

In addition, in the method of measuring invalid news continuous blood glucose, the selectivity of sugar molecules can be improved by configuring the electrode of porous platinum. Materials such as Ascorbic acid (AA) or Acetaminophen (AP) having a very high oxidation rate (refer to 15 (see FIG. 1)) generate noise in the electrical signal because the electrochemical signal is higher than the oxidation rate. As the electrode contains porous platinum, materials such as AA and AP (refer to 15, FIG. 1) having a high reaction rate react only on the electrode surface as shown in FIG. 1(a), and react on the surface of the pores 14a. On the other hand, glucose 16, which has a slow reaction rate, allows a reaction to occur on the surface of the pore 14a as shown in FIG. 1(b), so that the electrochemical signal from the glucose 16 can be amplified. There will be. In FIG. 1, for convenience of explanation, it is illustrated that the porous platinum layer has a uniform protrusion shape, but the present invention is not limited thereto. For example, the porous platinum layer may have a structure in which porous platinum particles are sprayed and stacked.

The blood glucose measurement system 1 may include a blood glucose measurement device 10 and a machine learning device 60.

The blood glucose measurement device 10 may measure a blood glucose value from the measurement current.

The blood glucose measurement device 10 may estimate or predict blood glucose from the continuous current signal measured from the invalid news electrode 12.

The blood glucose measurement device 10 may include at least one electrode 12 invading the skin and a voltage source (not shown) that applies a constant voltage between the at least one electrode 12.

At least one electrode 12 is formed in a needle shape, and a porous structure layer may be formed on the surface thereof. The pores 14a of the porous structure layer may be applied on a nanoscale. The porous structure layer may include a porous platinum layer 14, platinum, Pt. The porous platinum layer 14 may be formed on the surface of the electrode 12. The porous platinum layer 14 oxidizes glucose by contacting the glucose leaked from the capillary blood vessel (c).

Since the body of the needle-shaped electrode 12 invades under the skin (s), it may be formed of a material having strong acid resistance in order to prevent corrosion by interstitial fluid. The body of the electrode 12 may include at least one metal of gold (Au), silver (silver, Ag), stainless steel, and carbon (C). Therefore, it is possible to minimize the stimulation of pain points, thus minimizing user's rejection.

The voltage source is electrically connected to the electrode 12 to apply a constant voltage to the skin tissue invaded by the needle. In one example, the voltage source applies a voltage in any one of a direct current (DC) voltage, an alternating current (AC) voltage, or a voltage in which a direct current voltage and an alternating voltage are superimposed.

When a voltage is applied, the ionized glucose moves from the positive potential to the -potential direction by the applied voltage to form an electric current and is reduced at the electrode having the potential. It can be seen that the greater the amount of glucose in the blood, the greater the amount of ionized glucose, the greater the current value. Therefore, by measuring the amount of current flowing when a constant voltage is applied to the electrode 12 invading the skin, the amount of blood glucose can be calculated.

The blood glucose measurement device 10 may include a measurement unit 20 and a current characteristic search unit 30.

The measurement unit 20 is provided to obtain a current signal obtained by applying a voltage to the electrode 12. The measurement unit 20 may input a continuous current signal measured from the electrode 12. The measurement unit 20 may remove noise from the current signal measured from the electrode 12. The measurement unit 20 may convert an analog current signal into a digital current signal (A, see FIG. 3) through an ADC (Analog Digital Converter).

The current characteristic search unit 30 may extract features from digital current signals (hereinafter “current signals”) converted by the measurement unit 20.

The current characteristic search unit 30 may extract data corresponding to certain characteristics from the current signal (A, see FIG. 3) obtained by the measurement unit 20.

3 is a view showing a characteristic value in a measured current graph of a blood glucose measurement device according to an embodiment of the present invention.

The current characteristic search unit 30 may include at least one characteristic extraction unit. The current characteristic search unit 30 may include first to second feature extraction units 31 and 32.

The first feature extractor 31 may extract a first feature F1, which is an actual raw current at a measurement point (operation time). That is, the first characteristic F1 is the actual current value (raw current) at the measurement point (operation time) in the current signal (A). The measurement time may be a time point measured by the blood glucose measurement device 10 by a preset period, or may be a time point at which the user manipulates the blood glucose measurement device 10 to know the measured value.

The second feature extraction unit 32 extracts a second feature F2, which is a correction current at a measurement time, from a correction curve B generated based on the current signal A. can do.

The correction curve B is generated from the current signal A and may be formed by a certain equation.

Any one of the following equations may be applied to the correction curve.

Correction curve (1) R= -C ₁ /t+C ₂

Calibration curve (2) R=C ₃ e^(-t)+C ₄

Calibration curve (3) R=-C ₅ log(t)+C ₆

As for the correction curve B, as shown in FIG. 3, an equation formed within a predetermined error range with the current signal A may be applied.

When the correction curve B is specified, the correction current at the time of measurement can be specified. The second feature extraction unit 32 specifies any one correction curve within a certain error range among the correction curves (1) to (3), and the unknowns of the correction curve (C in the case of the correction curve (1)) ₁ , C ₂ , C ₃ and C ₄ for the correction curve 2, and C ₅ and C _{6 for the} correction curve 3) can be specified. The second feature extraction unit 32 may extract a correction current at the time of measurement from the specified correction curve B.

However, unlike this, the current characteristic search unit 30 may further include a correction curve generator (not shown). The correction curve generator may specify the current signal A and at least one of the correction curves 1 to 3 formed within a predetermined error range. In the case of the C _1, C _2, the calibration curve (2) for generating the calibration curve portion of the unknowns in the calibration curve (calibration curve (1) in the case of C _3, C _4, the calibration curve 3 is C ₅ , C ₆ ) can be specified. When the current characteristic search unit 30 includes a correction curve generator, the second feature extractor 32 may extract a correction current at the time of measurement from the correction curve specified by the correction curve generator. have.

The current characteristic search unit 30 may further include third and fourth feature extraction units 33 and 34.

The third feature extractor 33 may extract a third feature F3 that is an operation time. The measurement time may be calculated from the measurement start time (a point in which the time of FIG. 3 is 0) at which the current is generated as a starting point.

The fourth feature extraction unit 34 may extract a fourth feature F4, which is a total amount of electric charge up to an operation time. In the current signal A, the total amount of charge may be the area of the current signal A from the start point of the current signal to the measurement point as shown in FIG. 3.

The blood glucose measurement device 10 may include a blood glucose measurement unit 40.

The blood glucose measurement unit 40 is provided to measure blood glucose based on the generated machine learning model.

The blood glucose measurement unit 40 may input the feature values F1 to F4 extracted by the current characteristic search unit 30 and are calculated by the generated machine learning model to measure the blood glucose level. The machine learning model may be generated by the machine learning device 60 to be described later. The machine learning model is generated by the machine learning device 60 and stored in the blood glucose measurement device 10, so that the blood glucose measurement unit 40 can measure the blood glucose level through the stored machine learning model.

However, the present invention is not limited thereto, and the machine learning model may be stored in the machine learning apparatus 60. In this case, the blood glucose measurement unit 40 transmits the feature values F1 to F4 at the time of measurement through the communication unit to the machine learning device 60, and calculates the generated machine learning model by the machine learning device 60. It is possible to receive the blood glucose level.

The blood glucose measurement device 10 may include a display unit (not shown). The display unit is provided to output the blood glucose value measured by the blood glucose measurement unit 40. The display unit may be provided in the blood glucose measurement device 10, or may be transmitted to a separate user terminal 5 (refer to FIG. 2) to view the blood glucose level through the user terminal 5.

The user terminal 5 is not particularly limited as long as it has a communication function connected to the blood glucose measurement device 10 and a display function capable of outputting an image or text. For example, the user terminal 5 is a desktop computer, a laptop computer, a tablet PC, a wireless phone, a mobile phone, and a smart phone. , Smart watch, smart glass, e-book reader, portable multimedia player (PMP), portable game console, navigation device, digital camera, digital multimedia broadcasting (DMB) player , Digital audio recorder, digital audio player, digital picture recorder, digital picture player, digital video recorder, digital video player digital video player), but is not limited thereto. Here, the user may include a doctor as an expert and a measurer who uses the blood glucose measurement device 10. Users can include people or organizations that have access to the device.

The blood glucose measurement device 10 may include a communication unit (not shown). The communication unit may transmit and receive data such as a control signal or blood glucose value with the user terminal 5 or the machine learning device 60. The communication unit may transmit the measured blood glucose value to the user terminal 5 or the machine learning device 60 through wired or wireless communication. As an example, the communication unit is a user terminal 5 or machine learning device 60 and SWAP (Shared Wireless Access Protocol), Bluetooth, IrDA (Infrared Data Association), Zigbee (zigbee), Wi-Fi (wifi), NFC (Near Field Communication), cellular, and LTE (Long Term Evolution) communication protocols can be communicated through at least one of the communication protocols.

4 is a flowchart of a blood glucose measurement apparatus according to an embodiment of the present invention.

The measurement unit 20 may obtain a current signal A by applying a voltage to the electrode 12 (S110).

The current characteristic search unit 30 may extract feature values F1 to F4 from the continuous current signal A input to the measurement unit 20 (S120). The feature values may include a first feature, which is a current value at the time of measurement, and a second feature, which is a corrected current value corresponding to the time of measurement in the correction curve B generated from the obtained continuous current signal A. In addition, the feature values may further include a third feature that is a measurement time point and a fourth feature that is a total charge amount up to the measurement time point.

The blood glucose measurement unit 40 may measure the blood glucose value by inputting the feature values extracted from the current characteristic search unit 30 into the generated machine learning model (S130).

5 is a block diagram of a machine learning device of a blood glucose measurement system according to an embodiment of the present invention, and FIG. 6 is a diagram for correcting a time delay of the machine learning apparatus of a blood glucose measurement system according to an embodiment of the present invention.

The machine learning apparatus 60 may obtain current signal data and actual blood glucose measurement data from a database and learn about their relationship. The machine learning device 60 may generate a machine learning model for a relationship between the current and the actual blood glucose measurement value from the current data and the actual blood glucose measurement value. The database includes first data in which a current signal A generated by applying a voltage to the electrode 12 on which the porous platinum layer 14 is formed is accumulated for a plurality of subjects, and the actual blood glucose values of the plurality of subjects. The second data may be accumulated. The machine learning model may learn the relationship between the current signal A and the actual blood glucose value from the first and second data.

The machine learning model may be generated by the machine learning device 60. Machine learning models can be fed back by new data. That is, the generated machine learning model may be improved through the update of the blood glucose measurement device 10, and the improved machine learning model from the machine learning device 60 through the communication unit of the blood glucose measurement device 10 May be received.

The machine learning device 60 may include a preprocessor 70.

The preprocessor 70 may remove noise from current signal data, which is the first data. The preprocessor 70 is provided to correct a time delay for the current signal A of the first data and the current signal for the actual blood glucose measurement value of the second data as shown in FIG. 6. The actual blood glucose measurement value can be applied to the blood glucose measurement value through blood collection.

The machine learning device 60 may include a current characteristic learning unit 80.

The current characteristic learning unit 80 may extract data corresponding to certain characteristics of the current signal. The current characteristic learning unit 80 may extract feature values from the continuous current signal preprocessed by the preprocessor 70.

The current characteristic learning unit 80 may extract features from digital current signals (hereinafter “current signals”) preprocessed by the preprocessor 70.

The current characteristic learning unit 80 may include at least one characteristic extracting unit. The current characteristic learning unit 80 may include first to second characteristic extracting units 81 and 82.

In the description, the feature values extracted from the current characteristic learning unit 80 will be described with reference to the drawing of the current signal of FIG. 3.

The first feature extractor 81 may extract a first feature F1 (refer to FIG. 3 ), which is an actual raw current at a learning time. That is, the first characteristic F1 is the actual current value (raw current) at the learning time in the current signal (A). The learning time point may be a time point measured by a preset period, or may be an arbitrary time point at which learning takes place. The learning point may be any moment in the continuous current signal.

The second feature extractor 82 extracts a second feature F2, which is a correction current at a learning time, from the correction curve B generated based on the current signal A. can do.

The correction curve (B, see FIG. 3) is generated from a current signal (A, see FIG. 3), and may be formed by a certain equation.

Any one of the following correction curves may be applied to the correction curve.

Correction curve (1) R= -C ₁ /t+C ₂

Calibration curve (2) R=C ₃ e^(-t)+C ₄

Calibration curve (3) R=-C ₅ log(t)+C ₆

As for the correction curve B, as shown in FIG. 3, an equation formed within a predetermined error range with the current signal A may be applied.

When the correction curve is specified, a correction current value at the time of learning may be specified. The second feature extraction unit 82 specifies any one correction curve within a certain error range among correction curves (1) to (3), and the unknowns of the correction curve (C for correction curve (1)) ₁ , C ₂ , C ₃ and C ₄ for the correction curve 2, and C ₅ and C _{6 for the} correction curve 3) can be specified. The second feature extractor 82 may extract a correction current at a measurement point from the specified correction curve B.

However, unlike this, the current characteristic learning unit 80 may further include a correction curve generator (not shown). The correction curve generator may specify the current signal A and at least one of the correction curves 1 to 3 formed within a predetermined error range. In addition, the correction curve generation unit includes the unknowns of the corresponding correction curve (C ₁ and C ₂ for the correction curve (1), C ₃ and C ₄ for the correction curve (2), and C _{5 for the} correction curve (3)) , C ₆ ) can be specified. When the current characteristic learning unit 80 includes a correction curve generating unit, the second feature extracting unit 82 may extract a correction current at the time of measurement from the correction curve specified by the correction curve generating unit. have.

The current characteristic learning unit 80 may further include third and fourth feature extraction units 83 and 84.

The third feature extraction unit 83 may extract a third feature F3, which is a learning time. The learning time may be calculated from the learning start time (a point in which the time in FIG. 3 is 0) where the current is generated as a starting point.

The fourth feature extractor 84 may extract a fourth feature F4, which is a total amount of charge up to a learning time. In the current signal A, the total amount of charge may be the area of the current signal A from the start point of the current signal to the learning point as shown in FIG. 3.

The machine learning device 60 may include a machine learning model generator 90.

The machine learning model generation unit 90 performs machine learning by applying an algorithm based on information stored in the database, and is provided to generate a related machine learning model. The machine learning model generation unit 90 is provided to learn a relationship between the characteristic values of the current signal extracted from the current characteristic learning unit 80 and the actual blood glucose value of the second data. Algorithms such as artificial neural networks (ANNs) can be applied to generate machine learning models. However, the types of algorithms for machine learning models are not limited. As an example, as an algorithm of a machine learning model, deep neural networks can be applied as a kind of artificial neural network.

7 is a diagram illustrating a machine learning model generation of a machine learning apparatus of a blood glucose measurement system according to an embodiment of the present invention.

A certain number of feature value arrays may be randomly extracted from data in which feature values extracted through the feature extracting units from a certain time before the learning time to the learning time are arranged. When this is referred to as one data set, it is possible to obtain a plurality of data sets by repeating this.

In a plurality of datasets, the dependency between feature values may be removed through normalization. In a plurality of datasets, all input variables may be normalized to a similar size in order to prevent a training time delay and a decrease in accuracy caused by a difference between the absolute values of each feature value. For example, when there are 4 types of feature values and 20 feature value arrays, each data set may have 80 input values.

The normalized input values show prediction accuracy of a certain level or higher in the learning model, but the accuracy may decrease when other new data is input. In order to prevent such overfitting, the number of input values may be reduced by randomly removing input values through a drop out layer process in the structure of an artificial neural network as part of a regularization method as shown in FIG. 7.

The machine learning model can learn the relationship between the remaining input values and actual blood glucose measurements.

The machine learning model generated through this process may be stored in the blood glucose measurement device 10 or in the machine learning device 60. When stored in the machine learning device 60, the blood glucose measurement device 10 transmits the feature values extracted by the current characteristic learning unit 80 to the machine learning device 60 through the communication unit, and receives the blood glucose measurement value. I can.

In the present embodiment, it has been described that the machine learning model is generated by the machine learning device 60, but the present invention is not limited thereto, and the machine learning model generator 90 is provided in the blood glucose measurement device 10 to generate a machine learning model. have.

The blood glucose measurement apparatus 10 may measure blood glucose through the blood glucose measurement unit 20 based on the machine learning model generated through this process.

8(a), 8(b), and 8(c) are diagrams showing the accuracy of a machine learning model of a machine learning device of a blood glucose measurement system according to an embodiment of the present invention.

Reference BG (blood glucose) is the actual blood glucose measurement value, and Estimated BG is the blood glucose measurement value measured from the generated machine learning model.

8A shows the machine learning model generation unit 90 after extracting the first and second features F1 and F2 from the first and second feature extraction units 81 and 82 in the current characteristic learning unit 80. This is a diagram showing the accuracy of the machine learning model created through the process.

8B shows the machine after extracting the first, second, and third features (F1, F2, F3) from the first, second, and third feature extraction units 81, 82, and 83 in the current characteristic learning unit 80. It is a diagram showing the accuracy of the machine learning model generated through the learning model generation unit 90.

8C shows the first, second, third, and fourth features F1, F2, and the first, second, third, and fourth feature extracting units 81, 82, 83, 84 in the current characteristic learning unit 80. A diagram showing the accuracy of the machine learning model generated through the machine learning model generation unit 90 after extracting F3 and F4).

The machine learning model in which the relationship between the first and second features F1 and F2 and the actual blood glucose value is learned by the current characteristic learning unit 80 may have the same reliability as in FIG. 8A. That is, it can have high accuracy by learning the first and second features (F1, F2) and actual blood glucose measurement values, and as shown in FIGS. 8(b) and (c), the first, second, and third features (F1) , F2, F3) or the first to fourth features (F1 to F4) and the actual blood glucose measurement value may have higher accuracy.

In addition, the machine learning model in which the relationship between the first and second features F1 and F2 and the actual blood glucose value is learned by the current characteristic learning unit 80 may have a low error rate, such as the error rates in FIG. 8A. . In addition, as shown in FIGS. It can have a lower error rate for the value.

In the above, specific embodiments have been illustrated and described. However, it is not limited only to the above-described embodiments, and those of ordinary skill in the technical field to which the invention pertains will be able to perform various changes without departing from the gist of the technical idea of the invention described in the following claims. .

1: blood glucose measurement system 10: blood glucose measurement device
12: electrode 14: platinum layer
20: measurement unit 30: current characteristic search unit
40: blood glucose measurement unit 60: machine learning device
70: preprocessing unit 80: current characteristic learning unit
90: machine learning model generator

Claims (10)

A measuring unit having a porous platinum layer and obtaining a current signal obtained by applying a voltage to an electrode provided in a form that is minimally invasive on the skin;
A current characteristic search unit for extracting characteristic values from the obtained current signal;
And a blood glucose measurement unit that measures a blood glucose value from the feature values extracted by the current characteristic search unit through the generated machine learning model. The method of claim 1,
The current characteristic search unit,
A first feature extracting unit for extracting a first feature, which is a current value at a measurement point, from the obtained current signal;
And a second feature extracting unit for extracting a second feature that is a correction current value corresponding to the measurement point from the correction curve generated based on the obtained current signal. The method of claim 2,
The correction curve satisfies any one of the following equations:
R= -C ₁ /t+C ₂
R=C ₃ e^(-t)+C ₄
R=-C ₅ log(t)+C ₆ The method of claim 2,
The current characteristic search unit,
A third feature extraction unit for extracting a third feature as the measurement point;
A blood glucose measurement apparatus further comprising a fourth feature extracting unit for extracting a fourth feature that is a total amount of electric charge up to the measurement point. The method of claim 1,
The machine learning model,
First data on the acquired current signal for a plurality of subjects and second data on actual blood glucose values of the plurality of subjects are learned from the accumulated database, and the relationship between the first and second data is learned. Blood glucose measurement device. The method of claim 1,
The porous platinum layer is a blood glucose measurement device configured to be formed on the electrode surface. The method of claim 1,
The electrode,
Blood glucose measurement device comprising; at least one needle electrode having the porous platinum layer formed on the surface and formed in a needle shape. A measuring unit that obtains a current signal obtained by applying a voltage to an electrode provided with a porous platinum layer in a form that is minimally invasive on the skin;
A current characteristic search unit for extracting characteristic values from the obtained current signal,
A blood glucose measurement device including a blood glucose measurement unit configured to measure a blood glucose value from feature values extracted by the current characteristic search unit through the generated machine learning model;
A blood glucose measurement system comprising a; machine learning device for generating the machine learning model. At least one needle electrode on which a porous platinum layer is formed is invaded into the skin and a voltage is applied to obtain a current signal,
Extracting feature values from the obtained current signal,
A continuous blood glucose measurement method that measures blood glucose values from extracted feature values using the generated machine learning model. The method of claim 9,
The feature values are:
A first feature that is the current value at the time of measurement;
A second characteristic of a correction current value corresponding to the measurement point in the correction curve generated based on the obtained current signal;
A third feature that is the measurement point;
Continuous blood glucose measurement method comprising a; a fourth characteristic of the total charge amount up to the measurement point.

Images