Method, device, and system for monitoring glucose in body fluid on basis of electrochemical signal analysis

Abstract

The present invention pertains to a device for monitoring glucose in a body fluid on the basis of electrochemical signal analysis, and provides a device and system for monitoring glucose, the device comprising a processor which: receives, over a predetermined period of time, a current signal generated by an oxidation-reduction reaction between a body fluid of a subject and a reagent, so as to generate data on the basis of the current signal; extracts at least one item of feature information from the data; and determines a glucose concentration in the body fluid on the basis of the feature information.

Description

Method, device, and system for monitoring glucose in body fluids based on electrochemical signal analysis

[0001] Embodiments of the present invention relate to a method, apparatus, and system for glucose monitoring using body fluids. Specifically, embodiments of the present invention relate to a glucose monitoring method, apparatus, and system utilizing an electrochemical signal containing information regarding the concentration of glucose in body fluids.

[0002] Biological sample analysis technology utilizing electrochemical biosensors is widely used to quantitatively measure the concentrations of various biomolecules, such as glucose, electrolytes, and proteins. In particular, glucose sensors for blood glucose measurement in diabetes management are representative electrochemical biosensors widely used in clinical settings due to their high accuracy and reliability.

[0003] Saliva holds great potential as a sample for such analyses. Compared to blood, it can be collected non-invasively, minimizing the burden on the patient and reducing the risk of infection during the collection process. Furthermore, saliva contains not only glucose but also components reflecting various physiological states, such as electrolytes, hormones, and proteins, making it a noteworthy sample for monitoring overall health. However, the presence of these diverse components presents challenges in selectively identifying target analytes or accurately measuring their concentrations.

[0004] Mathematical formula 1 below represents the Cottrell equation.

[0005]

[0006] Here, i(t) is the current (A) measured at time t, n is the number of moles of electrons transferred, F is the Faraday constant (96485 C / mol), A is the surface area of ​​the electrode (cm²), D is the diffusion coefficient (cm² / s), C is the concentration of reactants in the solution (mol / cm³), and t is the reaction time (s).

[0007] The Cottrell Equation of Equation 1 above describes the current signal under diffusion-dominated conditions obtained from electrochemical measurements and expresses the principle of decreasing inversely proportional to the square root of time. This indicates that as the concentration of reactants at the electrode surface decreases, the diffusion process limits the current, and that electron transfer between the electrode and the solution in an electrochemical system is closely related to changes in reactant concentration. The Cottrell Equation quantitatively expresses this process and is used to analyze the response rate and efficiency of the sensor, as well as variations in the diffusion coefficient.

[0008] While these theoretical models describe behavior under ideal conditions, actual saliva samples exhibit minute signal fluctuations due to individual physiological differences, ionic strength, pH, and the presence of various biomolecules. Therefore, the development of advanced signal processing techniques and feature extraction methods is necessary to overcome the challenges posed by low analyte concentrations and complex matrices within body fluids.

[0009] Embodiments of the present invention can provide a glucose monitoring device and method that more accurately predict blood glucose by utilizing an electrochemical signal containing information regarding the concentration of glucose in body fluids.

[0010] Embodiments of the present invention can provide a glucose monitoring device and method that predict blood glucose more precisely by utilizing an electrochemical signal containing information regarding the concentration of glucose in body fluids.

[0011] The embodiments of the present invention aim to provide a blood glucose prediction system that can effectively correct errors caused by various factors, such as the influence of interfering substances in body fluids and differences in biosensors occurring during mass production, by extracting characteristic information of raw signals, and can improve reproducibility through multidimensional analysis.

[0012] To solve the aforementioned problem, the present invention provides a glucose monitoring device based on electrochemical signal analysis in a body fluid, comprising a processor that receives a current signal generated by an oxidation-reduction reaction between a subject's body fluid and a reagent for a certain period of time, generates data based on said current signal, extracts at least one characteristic information from said data, and determines the glucose concentration in said body fluid based on said characteristic information.

[0013] The above data may be time series data in which the current signal tends to decrease inversely proportional to the square root of time.

[0014] The above time series data may be data in which temporal characteristics are reflected through interval windowing.

[0015] The processor can extract a windowing interval from the data in which the rate of change of current is less than the reference rate of change.

[0016] The processor can map the percentile values ​​of the time series data to a Poincaré map, calculate the Mahalanobis distance on the Poincaré map, and determine the glucose concentration in the body fluid based on the standard deviation of the derivative of the Mahalanobis distance.

[0017] The above processor can determine a higher glucose concentration in the body fluid as the standard deviation of the Mahalanobis distance derivative is smaller.

[0018] The processor can extract the entropy of the current signal during the windowing interval and determine the glucose concentration in the body fluid based on the entropy.

[0019] The above processor can determine the glucose concentration in the body fluid to be higher as the entropy is smaller.

[0020] The processor can extract a current signal in the initial section of the data where the rate of change of current is greater than or equal to a reference rate of change, and determine the glucose concentration in the body fluid based on the current signal in the initial section.

[0021] The above processor can determine the glucose concentration in the body fluid to be higher as the current signal in the initial section is larger.

[0022] The processor can extract a current signal in the latter part of the data where the rate of change of current is less than a reference rate of change, and determine the glucose concentration in the body fluid based on the current signal in the latter part.

[0023] The above processor can determine a higher glucose concentration in the body fluid as the current signal in the latter section is larger.

[0024] The processor can extract the accumulated current energy value of the current signal during the specified time period from the data and determine the glucose concentration in the body fluid based on the accumulated current energy value.

[0025] The above processor can determine a higher glucose concentration in the body fluid as the accumulated current energy value increases.

[0026] The processor can extract the rate of change of the first and second current derivative values ​​in the initial section of the data where the rate of change of the current is greater than or equal to the reference rate of change, and determine the glucose concentration in the body fluid based on the rate of change of the first and second current derivative values.

[0027] The above processor can determine a higher glucose concentration in the body fluid as the difference between the rates of change of the first and second current derivative values ​​increases.

[0028] The processor can extract the rate of change of the first and second current derivative values ​​in the latter part of the data where the rate of change of the current is less than the reference rate of change, and determine the glucose concentration in the body fluid based on the rate of change of the first and second current derivative values.

[0029] The above processor can determine a higher glucose concentration in the body fluid as the difference between the rates of change of the first and second current derivative values ​​increases.

[0030] The processor can extract the rate of change of the moving average line of the current signal from the data and determine the glucose concentration in the body fluid based on the rate of change of the moving average line.

[0031] The above processor can determine a higher glucose concentration in the body fluid as the rate of change of the above moving average line increases.

[0032] The processor can extract the standard deviation of the current signal in the windowing interval and determine the glucose concentration in the body fluid based on the standard deviation.

[0033] The above processor can determine the glucose concentration in the body fluid to be higher as the above standard deviation increases.

[0034] The processor can extract the skewness of the current signal in the windowing interval and determine the glucose concentration in the body fluid based on the skewness.

[0035] The above processor can determine a higher glucose concentration in the body fluid as the skewness is smaller.

[0036] The processor can extract the autocorrelation of the current signal in the windowing interval and determine the glucose concentration in the body fluid based on the autocorrelation.

[0037] The above processor can determine a higher glucose concentration in the body fluid as the autocorrelation is smaller.

[0038] The processor can extract the ratio of the power spectral density of the low frequency band to the high frequency band in the current signal during the aforementioned period of time, and determine the glucose concentration in the body fluid based on the ratio of the power spectral density.

[0039] The above processor can determine a higher glucose concentration in the body fluid as the power spectral density ratio increases.

[0040] The processor can extract the ratio of the power spectral density of the high frequency band to the low frequency band in the current signal during the windowing interval, and determine the glucose concentration in the body fluid based on the ratio of the power spectral density.

[0041] The above processor can determine a higher glucose concentration in the body fluid as the power spectral density ratio increases.

[0042] The processor can predict the blood sugar of the subject using at least one of a first model that predicts the blood sugar using linear regression with at least one characteristic information as an independent variable and the subject's blood sugar as a dependent variable, a second model machine-learned to predict the blood sugar based on at least one characteristic information, and a third model deep-learned to predict the blood sugar based on at least one characteristic information.

[0043] The glucose monitoring device further includes an interface into which a biosensor containing the reagent is inserted and a voltage application unit, and the processor can control the voltage application unit to apply voltage to the biosensor and receive the current signal from the biosensor.

[0044] The present invention provides a glucose monitoring system based on electrochemical signal analysis in a body fluid, comprising: a body fluid collector for collecting a subject's body fluid; a biosensor including a reagent; and a glucose monitoring device that applies voltage to the biosensor, receives a current signal generated by an oxidation-reduction reaction between the body fluid and the reagent, generates data based on the current signal, extracts at least one characteristic information from the data, and determines the glucose concentration in the body fluid based on the characteristic information.

[0045] The present invention provides a glucose monitoring method based on electrochemical signal analysis by a processor, comprising the steps of: receiving a current signal generated by an oxidation-reduction reaction between a subject's body fluid and a reagent for a certain period of time and generating data based on the current signal; extracting at least one characteristic information from the data; and determining the glucose concentration in the body fluid based on the characteristic information.

[0046] According to one embodiment of the present invention, accurate and precise blood glucose prediction is made possible by utilizing an electrochemical signal containing information regarding the concentration of glucose in body fluids.

[0047] According to one embodiment of the present invention, the accuracy of blood glucose prediction can be improved by incorporating an algorithm through a firmware update without the need to change the hardware structure of the glucose monitoring system, thereby increasing ease of use and economic efficiency.

[0048] According to one embodiment of the present invention, by extracting characteristic information of a signal, errors caused by various factors, such as the influence of interfering substances in body fluids and differences in biosensors occurring during mass production, can be effectively corrected, and it also helps to improve reproducibility through multidimensional analysis.

[0049] FIG. 1 is a schematic diagram illustrating a glucose monitoring system according to one embodiment of the present invention.

[0050] FIG. 2 is a block diagram illustrating the configuration of a glucose monitoring device according to one embodiment of the present invention.

[0051] Figure 3 is a box plot showing the experimental results of the relationship between the standard deviation of the Mahalanobis distance derivative and glucose concentration.

[0052] Figure 4 is a diagram illustrating the early and latter sections of time series data as examples.

[0053] Figure 5 is a box plot graph showing the experimental results of the relationship between the current value and glucose concentration in the initial section.

[0054] Figure 6 is a box plot showing the experimental results of the relationship between the current value and glucose concentration in the latter part of the section.

[0055] Figure 7 is a graph showing time series data and cumulative current energy according to two different current signals.

[0056] Figure 8 is a box plot graph showing the experimental results of the relationship between cumulative current value and glucose concentration.

[0057] Figure 9 is a box plot showing the experimental results of the relationship between current energy values ​​and glucose concentration.

[0058] Figure 10 is a graph showing time series data and an initial section according to two different current signals.

[0059] Figure 11 is a box plot showing the experimental results of the relationship between the rate of change of the current derivative value in the initial section and glucose concentration.

[0060] Figure 12 is a box plot graph showing the experimental results of the relationship between the rate of change of the current derivative value in the latter section and glucose concentration.

[0061] Figure 13 is a box plot showing the experimental results of the relationship between the rate of change of the moving average line and glucose concentration.

[0062] Figure 14 is a diagram illustrating an exemplary windowing interval in time series data.

[0063] Figure 15 is a box plot graph showing the experimental results of the relationship between standard deviation and glucose concentration.

[0064] Figure 16 is a box plot showing the experimental results of the relationship between skewness and glucose concentration.

[0065] Figure 17 is a box plot showing the experimental results of the relationship between autocorrelation and glucose concentration.

[0066] Figure 18 is a box plot graph showing the experimental results of the relationship between the first power spectral density ratio of the low frequency band to the high frequency band and glucose concentration over a certain period of time.

[0067] Figure 19 is a box plot showing the experimental results of the relationship between the second power spectral density ratio of the high frequency band to the low frequency band and glucose concentration during the windowing interval.

[0068] Figure 20 is a graph showing the rate of change of the current signal over time.

[0069] Figure 21 is a box plot showing the experimental results of the relationship between entropy and glucose concentration during the windowing period.

[0070] FIG. 22 is a diagram showing the operation of a glucose monitoring device according to the first embodiment of the present invention.

[0071] FIG. 23 is a diagram showing the operation of a glucose monitoring device according to a second embodiment of the present invention.

[0072] FIG. 24 is a diagram showing the operation of a glucose monitoring device according to a third embodiment of the present invention.

[0073] FIG. 25 is a schematic diagram illustrating a glucose monitoring framework according to one embodiment of the present invention.

[0074] Hereinafter, preferred embodiments according to the present invention will be described in detail with reference to the accompanying drawings. The detailed description disclosed below, together with the accompanying drawings, is intended to describe exemplary embodiments of the present invention and is not intended to represent the only embodiment in which the present invention can be practiced. In order to clearly explain the present invention in the drawings, parts unrelated to the description may be omitted, and the same reference numerals may be used for identical or similar components throughout the specification.

[0075] FIG. 1 is a schematic diagram illustrating a glucose monitoring system according to one embodiment of the present invention.

[0076] The glucose monitoring system (1) of FIG. 1 may include a body fluid collector (10), a biosensor (20), and a glucose monitoring device (100).

[0077] The body fluid collector (10) may consist of a body fluid collecting swab, a plunger connected to the swab, and a body fluid collecting barrel having a discharge port at the end into which the plunger is inserted. The body fluid collector (10) operates by discharging the body fluid through the discharge port after the body fluid is absorbed into the swab connected to the plunger, and then the swab is placed into the body fluid collecting barrel and the plunger is pushed to compress the swab. The body fluid collector (10) may further include a filter to remove interfering substances contained in the body fluid.

[0078] The biosensor (20) is inserted into the glucose monitoring device (100) and may be implemented in the form of a strip, but is not limited thereto. Additionally, the biosensor (20) may further include a reagent that causes an oxidation-reduction reaction after contact with the body fluid discharged from the body fluid collector (10).

[0079] A glucose monitoring device (100) is a device that determines the glucose concentration in a body fluid by applying voltage to a biosensor (20) and analyzing the current signal generated by the oxidation-reduction reaction between the body fluid and the reagent, and predicts blood glucose based on the glucose concentration. Here, the body fluid may include saliva and blood, etc.

[0080] The present invention proposes a glucose monitoring device (100) and a method thereof that predict blood glucose more accurately and with higher precision by analyzing an electric current signal obtained through body fluids using at least one model.

[0081] Hereinafter, the configuration and operation of a glucose monitoring device according to one embodiment of the present invention will be described in detail with reference to the drawings.

[0082] FIG. 2 is a block diagram illustrating the configuration of a glucose monitoring device according to one embodiment of the present invention.

[0083] A glucose monitoring device (100) according to one embodiment of the present invention may include an input unit (110), a communication unit (120), a display unit (130), a storage unit (140), a voltage application unit (150), an analog-digital converter (ADC) (160), and a processor (170).

[0084] The input unit (110) generates input data in response to user input of the glucose monitoring device (100). For example, the user input may be a user input that starts the operation of the glucose monitoring device (100), a user input requesting a blood glucose prediction, etc., and may also be applied without limitation if it is a user input necessary to predict blood glucose using body fluids.

[0085] The input unit (110) includes at least one input means. The input unit (110) may include a menu button, a keyboard, a key pad, a dome switch, a touch panel, a touch key, etc.

[0086] The communication unit (120) can perform communication with an external device, such as a server, to transmit and receive measurement data, information regarding multiple characteristics, a first model that predicts blood sugar using linear regression, a second model that predicts blood sugar using machine learning, a third model that predicts blood sugar using deep learning, blood sugar information, etc.

[0087] The communication unit (120) can perform wireless communication such as 5G (5th generation communication), LTE-A (long term evolution-advanced), LTE (long term evolution), Wi-Fi (wireless fidelity), Bluetooth, or wired communication such as LAN (local area network), WAN (Wide Area Network), power line communication, USB (Universal Serial Bus), HDMI (High-Definition Multimedia Interface), Display Port, etc. For wired communication, the communication unit (120) may include a connector that allows the glucose monitoring device (100) to be physically connected to an external device, for example, a USB port, an earphone jack terminal, an HDMI connector, or an SD card connector. In addition, the communication unit (120) may further include an interface into which a biosensor (20) can be inserted.

[0088] The display unit (130) displays display data according to the operation of the glucose monitoring device (100). The display unit (130) can display a screen displaying blood glucose, a screen receiving user input, etc.

[0089] The display unit (130) includes a liquid crystal display (LCD), a light emitting diode (LED) display, an organic light emitting diode (OLED) display, a micro electro mechanical systems (MEMS) display, and an electronic paper display. The display unit (130) can be combined with the input unit (110) to be implemented as a touch screen.

[0090] The storage unit (140) stores operation programs of the glucose monitoring device (100). The storage unit (140) includes storage with non-volatile properties that can preserve data (information) regardless of whether power is provided, and memory with volatile properties in which data to be processed by the processor (170) is loaded and data cannot be preserved if power is not provided. Storage includes flash memory, HDD (hard-disc drive), SSD (solid-state drive), ROM (Read Only Memory), etc., and memory includes buffer, RAM (Random Access Memory), etc.

[0091] The storage unit (140) can store raw signals, multiple characteristic information, a first model that predicts blood sugar using linear regression, a second model that predicts blood sugar using machine learning, a third model that predicts blood sugar using deep learning, blood sugar information, etc. The storage unit (140) can store computation programs, etc., that are necessary in the process of extracting information regarding characteristics, obtaining blood sugar information, and outputting the final blood sugar.

[0092] The voltage application unit (150) applies voltage to the biosensor (20) inserted into the glucose monitoring device (100) under the control of the processor (170). At this time, the applied voltage may be a constant direct current (DC) voltage.

[0093] The ADC (160) is configured to convert an analog signal received from the biosensor (20) into a digital signal according to the oxidation-reduction reaction between the body fluid and the reagent layer after the voltage is applied by the voltage application unit (150).

[0094] The processor (170) can execute software, such as a program, to control at least one other component (e.g., hardware or software component) of the glucose monitoring device (100) and can perform various data processing or operations.

[0095] A processor (170) according to one embodiment of the present invention receives a current signal generated by an oxidation-reduction reaction between a subject's body fluid and a reagent for a certain period of time, generates data based on the current signal, extracts at least one characteristic information from the data, and can determine the glucose concentration in the body fluid based on the characteristic information. Here, the certain period of time may be 10 seconds, but is not limited thereto.

[0096] Additionally, the processor (170) can predict the subject's blood sugar based on the determined glucose concentration.

[0097] Here, the processor (170) can predict the blood sugar of a subject using at least one of a first model that predicts blood sugar using linear regression with at least one characteristic information as an independent variable and the subject's blood sugar as a dependent variable, a second model machine-learned to predict blood sugar based on at least one characteristic information, and a third model deep-learned to predict blood sugar based on at least one characteristic information.

[0098] At this time, the processor (170) may train the first model, the second model and / or the third model, or use the first model, the second model and / or the third model that has already been trained and built by porting and debugging from an external source, and is not limited to any one of these.

[0099] Meanwhile, the processor (170) may perform at least some of the data analysis, processing, and result information generation for performing the above operations using at least one of machine learning, artificial neural network, or deep learning algorithms as a linear regression, rule-based, or artificial intelligence algorithm. Examples of neural networks may include models such as CNN (Convolutional Neural Network), DNN (Deep Neural Network), RNN (Recurrent Neural Network), and Multilayer Perceptron (MLP).

[0100] A processor (170) according to one embodiment of the present invention receives a current signal generated by an oxidation-reduction reaction between a subject's body fluid and a reagent for a certain period of time, generates data based on the current signal, and can extract at least one characteristic information from the data.

[0101] Here, the data may include a signal obtained through the biosensor (20) or values ​​obtained by utilizing it. The signal obtained through the biosensor (20) may include, for example, a raw signal, and the raw signal is an electrochemically responsive current signal containing information regarding the concentration of glucose in the body fluid obtained from the biosensor (20) according to the oxidation-reduction reaction between the body fluid and the reagent layer after voltage application. At this time, the processor (170) may amplify the responsive current signal using an amplifier, for example, an operational amplifier (Opamp).

[0102] The above data is time series data, and the processor (170) can perform preprocessing such as data cleaning and processing of missing values, and can normalize and use it if necessary.

[0103] The above data can be calculated by converting the collected Analog-to-Digital Converter (ADC) time series data into a current signal as shown in the following mathematical formula 2.

[0104]

[0105] Based on the above data, signal features can be extracted to correct errors caused by differences between interfering substances in body fluids and biosensors.

[0106] The glucose diffusion signal obtained from electrochemical measurements exhibits a characteristic of decreasing inversely proportional to the square root of time according to the Cottrell equation. That is, the current I(t) resulting from the oxidation-reduction reaction of the electrochemical biosensor is expressed in the form of the following mathematical equation 2.

[0107] In the above mathematical formula 2, V out is the output voltage of the amplifier circuit, V in(Offset voltage) represents the input voltage and offset voltage. The amplification resistor represents the value of the resistance used in the amplifier, which can determine the amplifier's gain. The larger the value of the amplifier resistor, the greater the amplification ratio of the output signal. I(t) sensor The responsive current signal from the biosensor (20) that changes over time (t) may be a signal generated as a result of an electrochemical reaction according to the glucose concentration in the saliva.

[0108] The above mathematical formula 2 is an electrochemical induction current signal (I(t) generated in the biosensor (20) sensor It is used to calculate ). Specifically, when voltage is applied to the biosensor (20) by the voltage application unit (150), the electrons generated by the oxidation-reduction reaction of glucose in saliva are transferred from the electron transfer medium to the electrode, and the resulting current signal can be converted into a digital signal through an ADC and processed as a raw signal.

[0109] The sampling rate for acquiring raw signals over time may be 60 Hz or higher. For example, 60 or more ADC signals may be acquired per second, and characteristic information may be extracted from the acquired signals and used for blood glucose prediction. Alternatively, the period for acquiring raw signals may preferably be 100 Hz, but the present invention is not limited thereto.

[0110] The resolution of the raw signal of a glucose monitoring device may be 10 nA or less. For example, if the resolution is 2 nA, the ADC signal can be measured with a minimum resolution of 2 nA. When the period for acquiring the raw signal is low (high sampling rate) or the resolution of the raw signal is high, precise raw signal data can be measured, and by using this to extract various characteristic information, the accuracy of blood glucose prediction can be improved.

[0111] According to the Cottrell equation of Equation 1 mentioned above, as time t elapses, the current decreases inversely proportional to the square root of time. While this theoretical model describes behavior under ideal conditions, in actual body fluids, minute signal fluctuations may occur due to individual physiological differences, viscosity, ionic strength, pH, and the presence of various biomolecules.

[0112] Feature extraction functions developed to identify the hidden characteristics of body fluid components are essential for quantifying and interpreting the complex properties of electrochemical signals. This enables the effective correction of errors caused by the influence of interfering substances in body fluids and production differences between biosensors, and also helps improve reproducibility through multidimensional analysis.

[0113] Information regarding characteristics is information regarding signal characteristics that can be extracted from measurement data, and may broadly include state space reconstruction information, time-domain information, variability information, frequency domain information, complexity information, etc.

[0114] Characteristic information according to one embodiment of the present invention may be extracted from a raw signal obtained by applying a constant DC voltage and measurement data based thereon.

[0115] State space reconstruction information may include quantified information based on Mahalanobis distances regarding the distribution of state variables extracted using Poincaré sections obtained from the raw signal. By utilizing the Hilbert Transform to reconstruct the raw signal, which is a time-series signal, into phase space, changes in the signal's instantaneous amplitude and phase can be dynamically analyzed. Here, instantaneous amplitude represents the signal strength and can be used to dynamically analyze changes in signal intensity over time. Phase represents the pattern of periodic signal change and can be used to describe the signal's relative position with respect to the time axis. By utilizing the amplitude and phase information obtained through the Hilbert Transform, time-series data expressed in a simple one-dimensional space can be transformed into a high-dimensional space called phase space. Transformation into phase space makes it possible to more clearly distinguish between the periodic and non-periodic components of the signal, which contributes to understanding the complex characteristics and dynamic patterns of the signal.

[0116] Poincaré maps can be utilized for the analysis of state-space reconstruction information. A Poincaré map is a technique that extends from a low-dimensional space to a high-dimensional space to visually analyze the complex characteristics of a signal. By representing changes in a time-series signal in the form of a trajectory, Poincaré maps enable a more intuitive understanding of the signal's non-linear characteristics, dynamic change patterns, and periodicity, allowing for a clearer and more intuitive grasp of the signal's characteristics. Furthermore, through Poincaré maps, it is possible to visually understand how a signal changes in complex situations or conditions.

[0117] The transformed signal can be transformed into an N-dimensional space. The time series signal transformed into an N-dimensional space forms a specific trajectory according to the change in position within the N-dimensional space, and this trajectory can represent the change in position and the rate of change of the signal.

[0118] In N-dimensional space, signal analysis is possible using a standardized rate of change that varies at a constant ratio, unaffected by signal magnitude or concentration. Through such standardization, standardized analysis of the signal becomes possible by utilizing the characteristic that the rate of change maintains a constant ratio regardless of high or low concentrations. This allows for the dynamic identification of how the signal moves and changes under specific concentration conditions, and enables the clear identification of consistent signal patterns or outliers. For example, if the trajectory of a signal moving within N-dimensional space exhibits a deviation from the general trend, the impact of interfering factors on the signal can be quantitatively evaluated. This analysis of deviations can be utilized to numerically verify the influence of interfering factors on the electron transport process by comparing it with the statistical spatial rate of change of the time-series signal. This provides important insights for evaluating how the sensitivity of a biosensor changes under specific conditions and for detecting or correcting for interfering factors.

[0119] Since there may be differences in glucose concentration or blood glucose levels depending on the user, applying standardization to the rate of change of glucose concentration can provide a standardized blood glucose prediction system for multiple users. The blood glucose prediction system according to an embodiment of the present invention can extract characteristic information by extracting signals at specific points in time from a time-series ADC signal and mapping them to a two-dimensional space. According to one embodiment of the present invention, when the signal intensity at time is given, a sequence (Pn) can be obtained from the following mathematical formula 3.

[0120]

[0121] In the above mathematical formula 3, n represents the sample number, t represents time, and I represents the current value. Pn represents a data point on the Poincaré cross-section and is converted into a normalized value using the ratio between two consecutive signals. For example, if the number of samples is 100, the sequence (Pn) can be obtained by substituting n=1, 2, 3, …, 100.

[0122] Specifically, (x, y) coordinates are drawn one by one within the tolerance ellipse as the value of n increases, and the Mahalanobis distance from the center of the tolerance ellipse to each (x, y) coordinate can be calculated in a time series. Subsequently, by differentiating each Mahalanobis distance with respect to time and calculating the standard deviation of the value, it can be seen that the standard deviation value tends to decrease as the glucose concentration increases.

[0123] Specifically, the processor (170) maps percentile values ​​of a time series signal to a Poincaré map, calculates a Mahalanobis distance from an N-dimensional median based on data mapped in the Poincaré map, and determines the glucose concentration in the body fluid based on the standard deviation of the derivative of the calculated Mahalanobis distance.

[0124] Here, the percentile values ​​of the time series signal are mapped onto a Poincaré map and converge to (0, 0), and the convergence region is formed by an ellipse, so that the signal values ​​are located within the ellipse. The Mahalanobis distance refers to the distance between each coordinate value from the center of the ellipse and reflects the covariance of each coordinate axis.

[0125] Figure 3 is a box plot showing the experimental results of the relationship between the standard deviation of the Mahalanobis distance derivative and glucose concentration. Here, the x-axis represents glucose concentration and the y-axis represents the standard deviation of the Mahalanobis distance derivative.

[0126] Referring to Figure 3, it can be seen that glucose concentration is higher when the standard deviation of the Mahalanobis distance derivative is smaller.

[0127] Accordingly, the processor (170) can determine the glucose concentration in the body fluid to be higher as the standard deviation of the Mahalanobis distance derivative is smaller.

[0128] Time domain information is information that can be analyzed through changes in variables over time, and may include induction current, electrical energy, the instantaneous rate of change (Differentiation) of the induction current signal, and the degree of change (Envelope) of the moving average line.

[0129] Following an electrochemical reaction, the current signal changes over time due to diffusion. By tracking this time-series information resulting from diffusion in the time domain, the dynamic patterns and characteristics of the current signal can be analyzed. For example, based on time-domain information, various data can be derived, such as the rate of decline of the current signal due to diffusion, the rate of change over time, or irregular signal patterns.

[0130] Time-domain information (e.g., sensitized current) has a constant correlation with blood glucose values; however, this correlation may vary among users due to physiological differences (interfering factors). Therefore, by comparing time-domain information—characteristic data extracted by measuring raw signals—it is possible to identify which current signal values ​​have a high correlation with a specific user's blood glucose. This allows for the derivation of an optimized analysis method for each user and improves measurement accuracy.

[0131] Specifically, the processor (170) can extract a current value in an initial section of time series data where the rate of change of current is greater than or equal to a first reference rate of change, and determine the glucose concentration in the body fluid based on the current value in the initial section.

[0132] Additionally, the processor (170) can extract current values ​​in the latter part of the time series data where the rate of change of current is less than the second reference rate of change, and determine the glucose concentration in the body fluid based on the current values ​​in the latter part.

[0133] Figure 4 is a diagram illustrating the early and latter sections of time series data as examples.

[0134] Referring to FIG. 4, the initial section (p1) may be a section between 0.01s and 0.1s where the rate of change of the current value is relatively large, and the latter section (p2) may be a section between 4.9s and 5s where the rate of change of the current value is relatively small, but is not limited thereto.

[0135] Figure 5 is a box plot showing the experimental results of the relationship between the current value and glucose concentration in the initial section. Here, the x-axis represents the glucose concentration, and the y-axis represents the current value in the initial section.

[0136] Referring to Figure 5, it can be seen that the glucose concentration is higher as the current value in the initial section increases.

[0137] Accordingly, the processor (170) can determine that the glucose concentration in the body fluid is higher as the current value in the initial section increases.

[0138] Figure 6 is a box plot showing the experimental results of the relationship between the current value and glucose concentration in the latter part of the section. Here, the x-axis represents the glucose concentration, and the y-axis represents the current value in the latter part of the section.

[0139] Referring to Figure 6, it can be seen that the glucose concentration is higher as the current value in the latter part of the section increases.

[0140] Accordingly, the processor (170) can determine that the glucose concentration in the body fluid is higher as the current value in the latter part of the section increases.

[0141] Figure 7 is a graph showing time series data and cumulative current energy according to two different current signals.

[0142] Referring to Fig. 7, even if the final current values ​​of two different current signals are the same, the accumulated current value or current energy value may be different.

[0143] Even if the two current values ​​measured after a sufficient amount of time have passed are the same, the signals changing over time may have been different. This may be due to differences caused by various factors, such as the ion distribution within the user's body and the user's physiological mechanisms.

[0144] Based on this, the processor (170) can extract an accumulated current value or current energy value over a certain period of time from the time series data and determine the glucose concentration in the body fluid based on the accumulated current value or current energy value.

[0145] Figure 8 is a box plot showing the experimental results of the relationship between cumulative current values ​​and glucose concentration. Here, the x-axis represents glucose concentration and the y-axis represents cumulative current values.

[0146] Referring to Figure 8, it can be seen that the glucose concentration is higher as the cumulative current energy value increases.

[0147] Figure 9 is a box plot showing the experimental results of the relationship between cumulative current energy values ​​and glucose concentration. Here, the x-axis represents glucose concentration and the y-axis represents current energy values.

[0148] Referring to Figure 9, it can be seen that the glucose concentration is higher as the cumulative current energy value increases.

[0149] Accordingly, the processor (170) can determine that the glucose concentration in the body fluid is higher as the accumulated current value or current energy value increases.

[0150] Figure 10 is a graph showing time series data and an initial section according to two different current signals.

[0151] Referring to Fig. 10, even if the final current values ​​of two different current signals are the same, the slope values ​​of the initial section (p1) may differ. That is, a difference in diffusion may occur.

[0152] Based on this, the processor (170) can extract the rate of change of the first current derivative value in the initial section (p1) of the time series data where the rate of change of the current is greater than or equal to the third reference rate of change, and determine the glucose concentration in the body fluid based on the rate of change of the first current derivative value. Here, the initial section (p1) may be a section between 0s and 1s, but is not limited thereto.

[0153] Figure 11 is a box plot showing the experimental results of the relationship between glucose concentration and the difference (slope value) between the rate of change of the first current derivative and the rate of change of the second current derivative. Here, the x-axis represents glucose concentration, and the y-axis represents the rate of change of the first current derivative and the rate of change of the second current derivative.

[0154] Referring to Figure 11, it can be seen that the glucose concentration is higher as the difference (slope value) between the rate of change of the first current derivative and the rate of change of the second current derivative is larger.

[0155] Here, the slope value can be calculated using the following mathematical formulas 4 and 5.

[0156]

[0157]

[0158] In the above mathematical formula 4, t nrepresents a specific point in time, and the subscript n represents the sample number of the raw signal. Equation 5 represents the difference between the two slope values ​​calculated through Equation 4. For example, if a current signal is sampled at a frequency of 100 Hz, 100 signals can be measured per second, and if the signals were measured for 10 seconds, from t1 to t 1000 There are samples of . The slope can be obtained through Equation 4. By substituting this into Equation 5, the difference in slope values ​​between non-adjacent time points can be calculated. In Equation 5, m and k are arbitrary natural numbers, which can be values ​​between 1 and 1000 based on the example above.

[0159] Accordingly, the processor (170) can determine a higher glucose concentration in the body fluid as the difference between the rates of change of the first and second current derivative values ​​increases.

[0160] Likewise, the processor (170) can extract the rate of change of the third and fourth current derivative values ​​in the latter part of the time series data where the rate of change of the current is less than the fourth reference rate of change, and determine the glucose concentration in the body fluid based on the rate of change of the third and fourth current derivative values. Here, the latter part may be the interval between 3s and 4s, but is not limited thereto.

[0161] Figure 12 is a box plot showing the experimental results of the relationship between the difference in the rate of change of the third and fourth current derivatives and glucose concentration. Here, the x-axis represents glucose concentration, and the y-axis represents the difference in the rate of change of the third and fourth current derivatives (slope value).

[0162] Referring to Figure 12, it can be seen that the glucose concentration is higher as the difference in the rate of change of the third and fourth current derivatives increases.

[0163] Accordingly, the processor (170) can determine a higher glucose concentration in the body fluid as the difference between the rates of change of the third and fourth current derivative values ​​increases.

[0164] Additionally, the processor (170) can extract the rate of change of the moving average line of current values ​​from the time series data and determine the glucose concentration in the body fluid based on the rate of change of the moving average line.

[0165] Figure 13 is a box plot showing the experimental results of the relationship between the rate of change of the moving average line and glucose concentration. Here, the x-axis represents glucose concentration and the y-axis represents the rate of change of the moving average line.

[0166] Referring to Figure 13, it can be seen that the glucose concentration appears higher as the rate of change of the moving average line increases.

[0167] Accordingly, the processor (170) can determine a higher glucose concentration in the body fluid as the rate of change of the moving average line increases.

[0168] Variation characteristic information may include standard deviation characteristic information, windowing skewness characteristic information, autocorrelation characteristic information, etc.

[0169] Standard deviation is information indicating how far the acquired raw signal deviates from the signal's mean value; it can be used to evaluate signal stability by quantifying the degree of signal dispersion. Windowing skewness is an indicator representing the asymmetry of the signal distribution; it allows for the analysis of the extent to which the raw signal is skewed larger or smaller than the mean within a specific interval, and this can be utilized to detect abnormal patterns or outliers in the signal. Autocorrelation is an indicator measuring how similar the raw signals are over time; this allows for the analysis of signal repeatability or periodicity and can be used to identify the signal's consistency and periodic characteristics.

[0170] Furthermore, based on variation characteristic information, it is possible to analyze the degree of instantaneous change in electron transfer during the oxidation-reduction reaction caused by the catalyst. For example, if the instantaneous change in electron transfer resulting from the oxidation-reduction reaction by the catalyst is large, it can be interpreted that the pH level is likely to deviate from neutral, or that there are many external interfering factors affecting the signal (e.g., interfering ions such as metal ions and salts). Through the quantitative analysis of external environmental factors based on such variation characteristic information, the influence of external environmental factors (e.g., pH changes, concentration of interfering ions, etc.) on the measurement results of the biosensor can be corrected or eliminated.

[0171] Specifically, the processor (170) can extract a windowing interval in the time series data where the rate of change of current is less than the fifth reference rate of change.

[0172] Figure 14 is a diagram illustrating an exemplary windowing interval in time series data.

[0173] Referring to FIG. 14, the windowing interval (p3) may be a interval between 0.2s and 5.5s where the rate of change of the current value is relatively small, but is not limited thereto.

[0174] Here, since the preceding section of the windowing section (p3) consists mostly of DC components and does not contain frequency components, signal analysis is difficult, and accuracy decreases when analyzing with DC components included. Therefore, only the windowing section (p3) is separated from the time series data, and variation characteristic information is extracted from the windowing section (p3).

[0175] Here, the processor (170) extracts the standard deviation of the current value in the windowing interval (p3) and determines the glucose concentration in the body fluid based on the standard deviation.

[0176] Figure 15 is a box plot showing the experimental results of the relationship between standard deviation and glucose concentration. Here, the x-axis represents glucose concentration and the y-axis represents standard deviation.

[0177] Referring to Figure 15, it can be seen that the glucose concentration appears higher as the standard deviation increases.

[0178] Accordingly, the processor (170) can determine a higher glucose concentration in the body fluid as the standard deviation increases.

[0179] Additionally, the processor (170) extracts the skewness of the current value in the windowing interval (p3) and determines the glucose concentration in the body fluid based on the skewness.

[0180] Figure 16 is a box plot showing the experimental results of the relationship between skewness and glucose concentration. Here, the x-axis represents glucose concentration and the y-axis represents skewness.

[0181] Referring to Figure 16, it can be seen that the glucose concentration is higher as the skewness decreases.

[0182] Accordingly, the processor (170) can determine a higher glucose concentration in the body fluid as the skewness decreases.

[0183] Additionally, the processor (170) extracts the autocorrelation of the current value in the windowing interval (p3) and determines the glucose concentration in the body fluid based on the autocorrelation.

[0184] Figure 17 is a box plot showing the experimental results of the relationship between autocorrelation and glucose concentration. Here, the x-axis represents glucose concentration and the y-axis represents autocorrelation.

[0185] Referring to Figure 17, it can be seen that glucose concentration is higher as autocorrelation decreases.

[0186] Accordingly, the processor (170) can determine a higher glucose concentration in the body fluid as the autocorrelation is smaller.

[0187] Frequency domain information provides important data for analyzing the frequency components of a specific signal and may include information such as Fast Fourier Transform (FFT) ratio characteristics. The FFT converts a time-domain signal into the frequency domain, enabling a clear analysis of the signal's frequency components and allowing for the quantification of signal strength in a specific frequency band. Such frequency analysis is essential for analyzing the performance characteristics of biosensors and detecting signals in specific frequency bands.

[0188] The Cottrell equation describes the diffusion current over time in electrochemical reactions and is typically used to model electron transfer reactions at electrode surfaces. While this equation describes the temporal variation of current when a DC voltage is applied, ideally, no frequency-based signals should exist. However, in actual biosensor systems, unexpected frequency components may be detected due to specific interfering substances present on the electrode surface or in the surrounding environment. These frequency components indicate that the system has been subjected to noise or external interference and provide important clues for analyzing the presence of interfering substances.

[0189] For example, if a signal in the 40Hz band of a biosensor exhibits a specific level of frequency power (dB), it may suggest the possibility that noise originating from the power supply has been introduced into the system. Power supply noise is a major cause of signal quality degradation in the external environment, and it is important to quantify and appropriately compensate for it. Additionally, if frequency power (dB) in the 2–5Hz range is detected, it can be interpreted as a phenomenon related to the charging and discharging of the capacitive double layer within the biosensor. Double layer capacitance is generated by the electric double layer formed at the interface between the electrode and the electrolyte and can fluctuate periodically depending on potential changes within the glucose monitoring system. Potential changes within the system occur because the sensor's resistive components change due to oxidation-reduction reactions. This phenomenon can serve as an important indicator for a more thorough analysis of the biosensor's electrochemical reaction characteristics.

[0190] Specifically, the processor (170) can extract a first power spectral density ratio of a low frequency band (e.g., 0.5 Hz to 2 Hz) to a high frequency band (e.g., 5 Hz to 20 Hz) for a certain period of time, and determine the glucose concentration in the body fluid based on the first power spectral density ratio.

[0191] Figure 18 is a box plot graph showing the experimental results of the relationship between the ratio of the first power spectral density of the low frequency band to the high frequency band and glucose concentration over a certain period of time. Here, the x-axis represents glucose concentration, and the y-axis represents the ratio of the first power spectral density of the low frequency band to the high frequency band over a certain period of time.

[0192] Referring to Figure 18, it can be seen that the glucose concentration is higher as the first power spectral density of the low frequency band relative to the high frequency band increases over a certain period of time.

[0193] Accordingly, the processor (170) can determine a higher glucose concentration in the body fluid as the first power spectral density ratio increases.

[0194] Additionally, the processor (170) can extract a second power spectral density ratio of a high frequency band (e.g., 5 Hz to 20 Hz) to a low frequency band (e.g., 0.5 Hz to 2 Hz) during the aforementioned windowing interval (p3), and determine the glucose concentration in the body fluid based on the second power spectral density ratio.

[0195] Figure 19 is a box plot showing the experimental results of the relationship between glucose concentration and the ratio of the second power spectral density of the high frequency band to the low frequency band during the windowing period. Here, the x-axis represents glucose concentration, and the y-axis represents the ratio of the second power spectral density of the high frequency band to the low frequency band over a certain period of time.

[0196] Referring to Fig. 19, it can be seen that the glucose concentration is higher as the second power spectral density of the high frequency band relative to the low frequency band increases during the windowing interval (p3).

[0197] Accordingly, the processor (170) can determine a higher glucose concentration in the body fluid as the second power spectral density ratio increases.

[0198] Complexity information utilizes various analytical techniques to quantitatively analyze the complex patterns and irregularities of a signal, with Sample Entropy and Approximate Entropy being representative indicators. These indicators are useful for detecting signal anomalies and identifying hidden dynamic patterns based on time-series data.

[0199] Sample entropy numerically represents the pattern complexity of a signal and measures the probability of similar patterns appearing within a given signal. A lower value indicates a signal that is regular and highly predictable, whereas a higher value indicates a signal that is more complex and irregular.

[0200] Approximate entropy is another method for evaluating signal variability or irregularity, quantifying pattern consistency within the signal. This metric is primarily useful for detecting changes in complex dynamic systems, such as the variability of physiological data.

[0201] These complexity analysis indicators can be utilized to analyze the redox reactions occurring in electron transfer mediators and the minute oscillations of current caused by ions. In glucose monitoring systems including biosensors, the complexity of electrical signals does not react sensitively to external environmental factors or changes in the glucose monitoring system. This is because the approach to patterns is probabilistic, and thus it reacts insensitively to external factors, much like looking at absolute values. For example, if there are many electron transfer processes due to the redox reaction of glucose, the complexity may tend to decrease. This tendency to decrease complexity may be influenced by any factor that causes the biosensor to deviate from its equilibrium state. For instance, if the oscillations of the current become simpler, it may indicate that the electron transfer process has been subject to interference (e.g., interference caused by the presence of glucose) and is not in an equilibrium state. Raw signal data may be expressed on the real axis, and performing a Hilbert transform on it can result in data expressed on the complex plane, including the imaginary axis. The instant frequency and instant phase of a signal can be calculated by applying the Hilbert transform to the raw signal, and the Hilbert transform can be expressed by the following mathematical equation 6.

[0202]

[0203] Figure 20 is a graph showing the rate of change of the current signal over time.

[0204] The rate of change of the signal can be measured by differentiating the signal data converted through the above mathematical equation 6 with respect to time. Here, the rate of change of the signal may refer to the rate of phase change or the instantaneous frequency, and when plotted over time, it is represented by a graph such as FIG. 20. In FIG. 20, the x-axis is the time axis, and the y-axis may refer to the rate of phase change or the instantaneous frequency of the original signal over time.

[0205] For example, the system can operate by releasing electrons as glucose is oxidized by oxidase or glucose dehydrogenase, and then transferring these electrons to an electrode via an electron transporter. In this process, the electrical signal generated changes in proportion to the glucose concentration, allowing for real-time measurement of glucose concentration. During the electron transport process, the electron transporter undergoes repeated oxidation and reduction, exhibiting a charge-discharge effect similar to that of a capacitor. It was confirmed that the charge-discharge phenomena during the redox process, which were not detectable in the raw signal, can be identified from the phase change rate or instantaneous frequency obtained through the aforementioned signal processing. This information can be utilized as characteristic data serving as the basis for predicting glucose levels in saliva or the user's blood glucose.

[0206] Specifically, the processor (170) extracts the entropy of the current value during the windowing interval (p3) and determines the glucose concentration in the body fluid based on the entropy.

[0207] Here, the processor (170) can pattern the signal of the windowing interval (p3) and determine the entropy based on the diversity of the pattern. That is, it can be determined that the entropy increases when there are various patterns, and decreases when there are many identical patterns.

[0208] Figure 21 is a box plot showing the experimental results of the relationship between entropy and glucose concentration during the windowing period. Here, the x-axis represents glucose concentration, and the y-axis represents entropy during the windowing period (p3).

[0209] Referring to Figure 21, it can be seen that the glucose concentration is higher when the entropy is smaller during the windowing period (p3).

[0210] Accordingly, the processor (170) can determine a higher glucose concentration in the body fluid as the entropy decreases.

[0211] A processor (170) according to one embodiment of the present invention can obtain blood glucose information using at least one of a first model using linear regression in which each of a plurality of characteristics is an independent variable and blood glucose is a dependent variable, a second model machine-learned to calculate blood glucose based on information regarding a plurality of characteristics, and a third model deep-learned to calculate blood glucose based on information regarding a plurality of characteristics.

[0212] The processor (170) can obtain blood glucose information using at least one of the first model, the second model, and the third model. For example, the processor (170) can obtain blood glucose information using the first model. As another example, the processor (170) can obtain blood glucose information using the first model and the second model. As yet another example, the processor (170) can obtain blood glucose information using the first model and the third model.

[0213] The processor (170) can obtain blood glucose information according to each characteristic (hereinafter referred to as the first blood glucose information) by using a first model created for each characteristic. For example, blood glucose information according to the current characteristics of the measurement data can be obtained by using linear regression with the current characteristics as the independent variable and blood glucose as the dependent variable. The processor (170) can represent the distribution of blood glucose corresponding to multiple characteristics as a probability density function (PDF).

[0214] The processor (170) can obtain corresponding blood glucose information (hereinafter referred to as second blood glucose information) by inputting information regarding multiple characteristics into the second model. Alternatively, the processor (170) can obtain corresponding blood glucose information (hereinafter referred to as third blood glucose information) by inputting information regarding multiple characteristics into the third model.

[0215] A processor (170) according to one embodiment of the present invention can output the final blood glucose of a subject using at least one blood glucose information.

[0216] FIG. 22 is a diagram showing the operation of a glucose monitoring device according to the first embodiment of the present invention.

[0217] Referring to FIG. 22, first, the glucose monitoring device (100) can collect measurement data (S410).

[0218] The glucose monitoring device (100) can extract characteristic information using measurement data (S420). The measurement data can be used as is, but in addition, the measurement data can be preprocessed and used. The process of preprocessing the measurement data (A) is described with reference to FIG. 22.

[0219] The glucose monitoring device (100) can input extracted characteristic information into a model (S430). The model may include a first model (401) using linear regression, a second model (402) using machine learning, and a third model (403) using deep learning. The characteristic information may also be used as is, either in whole or in part, but may also be used after preprocessing in whole or in part. The process (B) of preprocessing the characteristic information is described with reference to FIG. 23.

[0220] The glucose monitoring device (100) can predict the final blood glucose level using blood glucose information obtained through at least one model (S440).

[0221] Hereinafter, FIGS. 22 and FIGS. 23 illustrate each embodiment of processes A and B, and the glucose monitoring device (100) can be implemented to selectively perform each process.

[0222] FIG. 23 is a diagram showing the operation of a glucose monitoring device according to a second embodiment of the present invention.

[0223] Figure 23 describes the process (A) of preprocessing measurement data, as previously described with reference to Figure 22.

[0224] After collecting measurement data, the glucose monitoring device (100) can determine whether there are any abnormal values ​​in the measurement data (S411).

[0225] If an outlier is determined in the measurement data (Yes in S411), the glucose monitoring device (100) can collect the measurement data again through re-measurement (S412).

[0226] The glucose monitoring device (100) can determine whether there are abnormal values ​​based on whether the signal shape is abnormal, whether the sample is suitable, etc., and the determination method can use a physics-based method or a statistics-based method.

[0227] Cases judged to be outliers include, for example, cases where the raw signal does not exhibit a Cottrell curve, or cases where the waveform is modulated due to the transient response of the glucose monitoring device (100). For example, cases where a Cottrell curve is not exhibited may include cases where there is no reduction in current due to diffusion.

[0228] If no outlier is detected (No of S411), the glucose monitoring device (100) can perform temperature correction (S413). Temperature correction means correcting other temperature measurements to room temperature measurements. The glucose monitoring device (100) can correct other temperature measurements to room temperature measurements by utilizing the Cottrell equation and the Einstein-Stokes equation.

[0229] The glucose monitoring device (100) can perform an outlier determination on the temperature correction result after temperature correction (S414). If an outlier is determined as a result of the temperature correction (Yes in S414), the glucose monitoring device (100) can perform temperature correction again. If no outlier is determined as a result of the temperature correction (No in S414), the process proceeds to the characteristic information extraction step.

[0230] According to one embodiment of the present invention, more accurate characteristic information can be extracted by checking for outliers in the measurement data and correcting for temperature.

[0231] FIG. 24 is a diagram showing the operation of a glucose monitoring device according to a third embodiment of the present invention.

[0232] Figure 24 describes the process (B) of preprocessing feature information, as previously described with reference to Figure 23.

[0233] After extracting characteristic information, the glucose monitoring device (100) can determine whether the characteristic information is an outlier (S421).

[0234] The glucose monitoring device (100) can determine whether there are abnormalities based on whether the characteristic information has statistical abnormalities, whether the sample is suitable, etc., and the determination method can use a physics-based method or a statistics-based method.

[0235] If an outlier is determined in the characteristic information (Yes of S421), the glucose monitoring device (100) can re-collect the raw signal or re-extract the characteristic information through re-measurement.

[0236] If no outlier is detected (No. of S421), the process proceeds to the feature information extraction step.

[0237] Meanwhile, we will examine in detail cases where outliers are not detected. Criteria for identifying outliers can be established by utilizing the Signal Quality Index (SQI), which is defined through physics-based and statistical-based methods. This index plays a crucial role in evaluating signal quality to maintain system reliability and data accuracy.

[0238] In such cases, there may be instances where characteristic information is not judged to be an outlier but falls within the predefined range of signal quality indicators. For instance, this occurs when characteristic information exceeds a threshold but barely passes it. In such cases, it may be difficult to assess the reliability of the relevant characteristic information. For instance, it is difficult to distinguish whether the signal is distorted due to a performance issue with the biosensor or caused by the characteristics of the measurement target, such as bodily fluids.

[0239] In this case, the glucose monitoring device (100) can adjust the weights of the characteristic information. By adjusting the weights of the blood glucose prediction, the glucose monitoring device (100) can increase the consistency and reproducibility of the system. Adjusting the weights can contribute to reducing data uncertainty by adjusting the proportion that specific signal characteristics occupy in the overall analysis. This is effective in uncertainty situations where the variability of signal quality is high and it is difficult to predict all possible cases in advance, and a probabilistic approach can be applied to solve these problems. A probabilistic approach can contribute to deriving more reliable results by interpreting data uncertainty based on statistical distributions and patterns.

[0240] The glucose monitoring device (100) can perform a final blood glucose prediction by reflecting weights in the acquired blood glucose information.

[0241] FIG. 25 is a schematic diagram illustrating a glucose monitoring framework according to one embodiment of the present invention.

[0242] Referring to FIG. 25, the glucose monitoring framework (1100) may further include a server (200) and a user terminal (300) in addition to the glucose monitoring system (1).

[0243] The server (200) is a device that communicates with the glucose monitoring system (1) and performs at least one of a series of operations to monitor glucose from measurement data, and can be implemented in addition to the server, such as a cloud or a computer.

[0244] For example, the server (200) may perform at least one of the following operations: extracting information regarding a plurality of characteristics from data measured from the subject's body fluid; obtaining blood glucose information from a first model using linear regression with at least one characteristic as an independent variable and blood glucose as a dependent variable; obtaining blood glucose information from a second model machine-learned to calculate blood glucose based on information regarding a plurality of characteristics; obtaining blood glucose information from a third model deep-learned to calculate blood glucose based on information regarding a plurality of characteristics; and obtaining the subject's final blood glucose using at least one obtained blood glucose information.

[0245] The user terminal (300) is a terminal that displays data obtainable during the process of monitoring final blood glucose and other glucose obtained through the glucose monitoring system (1) and / or server (200), and can be implemented as a computer, smartphone, tablet PC, smart pad, laptop, etc. The user terminal (300) can store, for example, an application implemented to monitor glucose, and can display data received through the glucose monitoring device (100) and / or server (200) according to the operation of the application.

[0246] The glucose monitoring device and system of the present invention can be used in various industrial fields, such as biosensors.

Claims

1. In a glucose monitoring device in body fluids based on electrochemical signal analysis, A current signal generated by the oxidation-reduction reaction between the subject's body fluid and a reagent is received as input for a certain period of time, and data is generated based on the said current signal, and Extract at least one feature information from the above data, and A processor for determining the glucose concentration in the body fluid based on the above characteristic information Glucose monitoring device.

2. In Paragraph 1, The above data Time series data in which the current signal tends to decrease inversely proportional to the square root of time Glucose monitoring device.

3. In Paragraph 2, The above time series data Data that reflects temporal characteristics through interval windowing Glucose monitoring device.

4. In Paragraph 2, The above processor Extracting windowing intervals from the above data where the current change rate is less than the reference change rate Glucose monitoring device.

5. In Paragraph 2, The above processor Map the percentile values ​​of the above time series data to a Poincaré map, and Calculate the Mahalanobis distance from the above Poincaré map, and Determining the glucose concentration in the body fluid based on the standard deviation of the above Mahalanobis distance derivative Glucose monitoring device.

6. In Paragraph 5, The above processor The smaller the standard deviation of the above Mahalanobis distance derivative, the higher the glucose concentration in the above body fluid is determined. Glucose monitoring device.

7. In Paragraph 4, The above processor Extract the entropy of the current signal during the windowing interval, and Determining the glucose concentration in the body fluid based on the above entropy Glucose monitoring device.

8. In Paragraph 7, The above processor The smaller the entropy, the higher the glucose concentration in the body fluid is determined. Glucose monitoring device.

9. In Paragraph 2, The above processor Extract the current signal from the initial section of the above data where the rate of change of current is greater than or equal to the reference rate of change, and Determining the glucose concentration in the body fluid based on the current signal of the aforementioned initial section Glucose monitoring device.

10. In Paragraph 9, The above processor The higher the current signal in the aforementioned initial section, the higher the glucose concentration in the body fluid is determined. Glucose monitoring device.

11. In Paragraph 2, The above processor Extract the current values ​​from the above data for the latter part of the section where the rate of change of current is less than the reference rate of change, and Determining the glucose concentration in the body fluid based on the current value of the latter section above Glucose monitoring device.

12. In Paragraph 11, The above processor The higher the current signal in the latter section mentioned above, the higher the glucose concentration in the body fluid is determined. Glucose monitoring device.

13. In Paragraph 2, The above processor Extract the accumulated current energy value of the current signal for the aforementioned period of time from the above data, and Determining the glucose concentration in the body fluid based on the above accumulated current energy value Glucose monitoring device.

14. In Paragraph 13, The above processor The higher the above accumulated current energy value, the higher the glucose concentration in the above body fluid is determined. Glucose monitoring device.

15. In Paragraph 2, The above processor Extract the rate of change of the first and second current derivative values ​​in the initial section where the rate of change of current is greater than or equal to the reference rate of change from the above data, and Determining the glucose concentration in the body fluid based on the rate of change of the first and second current derivative values. Glucose monitoring device.

16. In Paragraph 15, The above processor The larger the difference in the rate of change between the first and second current derivative values, the higher the glucose concentration in the body fluid is determined. Glucose monitoring device.

17. In Paragraph 2, The above processor Extract the rate of change of the first and second current derivative values ​​in the latter part of the data where the rate of change of current is less than the reference rate of change, and Determining the glucose concentration in the body fluid based on the rate of change of the first and second current derivative values. Glucose monitoring device.

18. In Paragraph 17, The above processor The larger the difference in the rate of change between the first and second current derivative values, the higher the glucose concentration in the body fluid is determined. Glucose monitoring device.

19. In Paragraph 2, The above processor Extract the rate of change of the moving average line of current values ​​from the above data, and Determining the glucose concentration in the body fluid based on the rate of change of the moving average line. Glucose monitoring device.

20. In Paragraph 19, The above processor The greater the rate of change of the above moving average line, the higher the glucose concentration in the above body fluid is determined. Glucose monitoring device.

21. In Paragraph 20, The above processor Extract the standard deviation of the current signal in the above windowing interval, and Determining the glucose concentration in the body fluid based on the above standard deviation Glucose monitoring device.

22. In Paragraph 21, The above processor The larger the above standard deviation, the higher the glucose concentration in the above body fluid is determined. Glucose monitoring device.

23. In Paragraph 4, The above processor Extract the skewness of the current signal in the above windowing interval, and Determining the glucose concentration in the body fluid based on the above skewness Glucose monitoring device.

24. In Paragraph 23, The above processor The smaller the skewness, the higher the glucose concentration in the body fluid is determined. Glucose monitoring device.

25. In Paragraph 4, The above processor Extract the autocorrelation of the current signal in the above windowing interval, and Determining the glucose concentration in the body fluid based on the above autocorrelation Glucose monitoring device.

26. In Paragraph 25, The above processor The smaller the above autocorrelation, the higher the glucose concentration in the body fluid is determined. Glucose monitoring device.

27. In Paragraph 2, The above processor Extract the ratio of the power spectral density of the low frequency band to the high frequency band from the current signal during the above specified time, and Determining the glucose concentration in the body fluid based on the above power spectrum density ratio Glucose monitoring device.

28. In Paragraph 27, The above processor The higher the power spectrum density ratio, the higher the glucose concentration in the body fluid is determined. Glucose monitoring device.

29. In Paragraph 4, The above processor During the above windowing interval, the ratio of power spectral density of the high frequency band to the low frequency band is extracted from the current signal, and Determining the glucose concentration in the body fluid based on the above power spectrum density ratio Glucose monitoring device.

30. In Paragraph 29, The above processor The higher the power spectrum density ratio, the higher the glucose concentration in the body fluid is determined. Glucose monitoring device.

31. In Paragraph 1, The above processor Predicting the blood sugar of the subject using at least one of a first model that predicts the blood sugar using linear regression with at least one characteristic information as an independent variable and the subject's blood sugar as a dependent variable, a second model machine-learned to predict the blood sugar based on the at least one characteristic information, and a third model deep-learned to predict the blood sugar based on the at least one characteristic information. Glucose monitoring device.

32. In Paragraph 1, An interface into which a biosensor containing the above reagent is inserted; and It further includes a voltage application unit, and The above processor is, The above voltage application unit is controlled to apply voltage to the biosensor, and Receiving the current signal from the above biosensor Glucose monitoring device.

33. In an electrochemical signal analysis-based glucose monitoring system in body fluids, A fluid collector that collects the subject's body fluids; A biosensor comprising a reagent; and A glucose monitoring device that applies voltage to the biosensor, receives a current signal generated by an oxidation-reduction reaction between the body fluid and the reagent, generates data based on the current signal, extracts at least one characteristic information from the data, and determines the glucose concentration in the body fluid based on the characteristic information. A glucose monitoring system including 34. A method for monitoring glucose in body fluids based on electrochemical signal analysis by a processor, A step of receiving a current signal generated by the oxidation-reduction reaction between the subject's body fluid and a reagent for a certain period of time and generating data based on the said current signal; A step of extracting at least one characteristic information from the above data; and Step of determining the glucose concentration in the body fluid based on the above characteristic information A glucose monitoring method including

Images