Blood glucose monitoring and early warning method

Abstract

The invention provides a blood glucose monitoring and early warning method. The blood glucose monitoring and early warning method comprises the steps of a, collecting a skin surface spectral signal I (lambda), b, collecting skin surface temperature T, humidity H, environment light intensity L and environment electromagnetic interference E parameters, c, preprocessing the spectral signal I (lambda), removing noise N (I) through wavelet transform, and d, calculating the blood glucose monitoring and early warning parameters. And the signal intensity is corrected through an ambient light intensity compensation model. According to the blood glucose monitoring and early warning method, a non-invasive acquisition mode combining near infrared spectrum and visible light spectrum is adopted, and a multi-modal regression model and a dynamic interference compensation technology are innovatively introduced, so that the problem that the precision of a sensor is influenced by skin temperature, humidity and environmental electromagnetic interference in the prior art is solved; the timeliness and the stability of signal purification are improved by utilizing preprocessing and optimization algorithms of spectral signals, such as wavelet transform, empirical mode decomposition and fast Fourier transform.

Description

Technical Field

[0001] The present invention relates to the technical field of blood sugar monitoring, and in particular to a blood sugar monitoring early warning method. Background Art

[0002] A blood glucose monitoring and early warning method is usually composed of a blood glucose sensor, a data acquisition module, a data processing module and an early warning module. The system collects the glucose concentration in the patient's blood in real time through a blood glucose sensor, and the data acquisition module converts the original signal output by the sensor into processable data and transmits it to the data processing module. The data processing module analyzes the blood glucose concentration based on a preset blood glucose safety range (e.g., 3.9-7.8mmo l / L) and a dynamic trend analysis algorithm. When the blood glucose concentration exceeds the safety range or the blood glucose change trend indicates that a sharp fluctuation may occur, the early warning module will remind the user by sound, vibration or mobile terminal notification to help the user take timely measures to reduce the health risks caused by hypoglycemia or hyperglycemia.

[0003] Although the system plays an important role in blood sugar management, it still has several shortcomings. First, the accuracy of the sensor may be affected by skin temperature, humidity and environmental electromagnetic interference, resulting in reduced accuracy of the monitoring data. Too frequent false alarms may cause users to have "warning fatigue" and reduce their sensitivity to warning information. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a blood sugar monitoring early warning method, which solves the problem that the accuracy of the sensor is affected by skin temperature, humidity and environmental electromagnetic interference, resulting in a decrease in the accuracy of the monitoring data. Too frequent false alarms may cause users to have "early warning fatigue" and reduce their sensitivity to early warning information.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solutions: a blood sugar monitoring and early warning method, comprising the following steps:

[0006] a. Collect the skin surface spectral signal I(λ), where λ is the spectral wavelength, combining the near-infrared spectrum and the visible light spectrum signal;

[0007] b. Collect skin surface temperature T, humidity H, ambient light intensity L and ambient electromagnetic interference E parameters;

[0008] c. Preprocess the spectral signal I(λ), including removing noise N(I) through wavelet transform and correcting the signal intensity through the ambient light intensity compensation model to obtain the purified spectral signal:

[0009] I ′ (λ)=I(λ)-N(I)+L c , where L cis the light intensity compensation coefficient;

[0010] d. Based on the T, H, L and E parameters, the comprehensive interference compensation coefficient C is calculated using the multimodal regression model f(T,H,L,E) f ;

[0011] e. The interference compensation coefficient C f Introduce the spectral signal correction formula:

[0012] I c (λ)=I ′ (λ)× ( 1-C f) ;

[0013] Get the corrected spectral signal I c (λ);

[0014] f. From I c (λ) extracts the spectral feature vector X = {x 1 ,x 2 ,…,x n}, combined with the user's historical blood sugar data vector Y = {y 1 ,y 2 ,…,y n} and individual biological parameters for feature matching;

[0015] g. Use dynamic trend analysis algorithm, combined with time series model TS(t) and neural network model to predict future blood sugar change trend:

[0016]

[0017] in is the predicted blood glucose concentration, ∈ is the error;

[0018] i. Combine the user's daily activity records, dietary intake data and physiological cycle parameters to calculate the accuracy probability of the warning through the Bayesian decision model P(W|A,C);

[0019] h. Determine whether the current and predicted blood sugar status is within the personalized blood sugar safety range [G min ,G max ], generating a preliminary warning signal W;

[0020] j. Determine whether to send a warning signal and its level based on the probability P(W) and the warning level threshold θ;

[0021] k. Send personalized blood sugar management suggestions M, including diet, exercise and medication adjustment plans, via vibration, sound or mobile phone notification according to the warning level:

[0022] M=f(W,user profile,activitylog);

[0023] I. Users can view blood sugar status, trend analysis and intervention suggestions through the mobile interface, and can upload feedback information and update model parameters.

[0024] Preferably, the spectral signal I(λ) collected in step a uses a dual-wavelength alternating sampling technique, combined with a fixed wavelength and an adjustable wavelength signal enhancement mode, to improve the sensitivity of the spectral data, and the compensation of the ambient light intensity L in step b is achieved by adaptively adjusting the model L c =g(L), which can dynamically adapt to changes in indoor and outdoor collection environments.

[0025] Preferably, the spectral signal purification in step c combines empirical mode decomposition and fast Fourier transform to optimize the denoising process, thereby improving the timeliness and stability of the purified signal, and the comprehensive interference compensation coefficient C in step d f It is trained using a convolutional neural network, and the model input includes temperature, humidity, light intensity and electromagnetic interference parameters.

[0026] Preferably, the individual biological parameters include body mass index, insulin sensitivity and long-term blood sugar control index, and the neural network model in step g is a hybrid neural network combined with a long short-term memory network, which is used to capture the correlation between time series and static features.

[0027] Preferably, the personalized blood sugar safety range in step h [G min ,G max ] can be dynamically optimized based on the user's physique, age and daily activity patterns. The Bayesian decision model input is combined with the historical warning false alarm rate to reduce the frequency of false alarms and improve the credibility of warnings.

[0028] Preferably, the warning signal is presented in three ways: visual chart, voice prompt and emergency vibration reminder. The user can set the priority by himself. The management suggestion M in step k is generated in combination with user feedback data and doctor's advice, and can be sent to the medical service platform through remote synchronization.

[0029] Preferably, the early warning method performs real-time data processing through an edge computing platform, and performs periodic big data analysis through a cloud server to improve algorithm stability. The early warning method is compatible with wearable devices, including smart bracelets, smart watches and independent blood glucose monitors, and supports linkage with other health monitoring devices.

[0030] Preferably, the early warning method adopts a dynamic parameter tuning mechanism, regularly evaluates the interference compensation effect, and automatically adjusts the acquisition and calculation parameters to adapt to environmental changes. The early warning method supports the generation of personalized health reports based on artificial intelligence, and provides users with weekly or monthly blood sugar change summaries and health recommendations.

[0031] The present invention provides a blood sugar monitoring and early warning method. It has the following beneficial effects:

[0032] The present invention innovatively introduces a multimodal regression model and dynamic interference compensation technology by combining the non-invasive acquisition method of near-infrared spectrum and visible light spectrum, thereby solving the problem in the prior art that the accuracy of the sensor is affected by skin temperature, humidity and environmental electromagnetic interference. The preprocessing and optimization algorithms of spectral signals, such as wavelet transform, empirical mode decomposition and fast Fourier transform, are used to improve the timeliness and stability of signal purification. The interference compensation coefficient is dynamically adjusted through a convolutional neural network, and combined with a long short-term memory network and time series analysis, the accuracy and reliability of blood sugar prediction are significantly enhanced. In addition, the present invention can dynamically optimize the personalized blood sugar safety range, effectively improve the adaptability of the monitoring method, and meet the physiological needs of different users.

[0033] The present invention introduces a Bayesian decision model and comprehensively considers the user's daily activity records, dietary intake data and physiological cycle parameters, thereby greatly reducing the frequency of false alarms and preventing users from ignoring important information due to "warning fatigue". BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0035] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] like Figure 1 As shown, the embodiment of the present invention provides a blood sugar monitoring and early warning method, comprising the following steps:

[0037] a. Collect the skin surface spectral signal I(λ), where λ is the spectral wavelength, and combine the near-infrared spectrum and visible light spectrum signals. The collected spectral signal uses dual-wavelength alternating sampling technology, combined with fixed wavelength and adjustable wavelength signal enhancement mode to improve the sensitivity of spectral data. The dual-wavelength alternating sampling technology of the collected spectral signal is further combined with optical filters and adaptive gain adjustment modules to dynamically optimize the signal intensity and noise ratio in different bands. The sampling process dynamically models the distribution characteristics of the near-infrared and visible spectra, generates real-time correction parameters for spectral acquisition, and ensures the accuracy and consistency of the spectral signal.

[0038] b. Collect skin surface temperature T, humidity H, ambient light intensity L and environmental electromagnetic interference E parameters. The compensation of ambient light intensity can dynamically adapt to changes in indoor and outdoor collection environments through an adaptive adjustment model. The ambient light intensity compensation model is combined with a brightness distribution detection method based on multi-region comparison to achieve rapid correction in the case of abnormal local light intensity. At the same time, through comparative analysis with historical environmental parameter data, the weight of the compensation model is dynamically adjusted to adapt to the variability of the user's environment;

[0039] c. Preprocess the spectral signal I(λ), including removing noise N(I) through wavelet transform and correcting the signal intensity through the ambient light intensity compensation model to obtain the purified spectral signal:

[0040] I ′ (λ)=I(λ)-N(I)+L c , where L c For the light intensity compensation coefficient, spectral signal purification combines empirical mode decomposition and fast Fourier transform to optimize the denoising process and improve the timeliness and stability of the purified signal

[0041] d. Based on the T, H, L and E parameters, the comprehensive interference compensation coefficient C is calculated using the multimodal regression model f(T,H,L,E) f , comprehensive interference compensation coefficient C f It is trained by convolutional neural network, and the model input includes temperature, humidity, light intensity and electromagnetic interference parameters;

[0042] e. The interference compensation coefficient C f Introduce the spectral signal correction formula:

[0043] I c (λ)=I ′ (λ)× ( 1-C f) ;

[0044] Get the corrected spectral signal I c(λ), the corrected spectral signal is further optimized through normalization processing and multi-scale feature fusion technology to ensure the sensitivity and robustness of the spectral signal in the subsequent feature extraction process, and the computational complexity of the signal is reduced through the feature redundancy reduction module.

[0045] f.From I c (λ) extracts the spectral feature vector X = {x 1 ,x 2 ,…,x n}, combined with the user's historical blood sugar data vector Y = {y 1 ,y 2 ,…,y n} and feature matching with individual biological parameters, including body mass index, insulin sensitivity, and indicators of long-term glycemic control;

[0046] g. Use dynamic trend analysis algorithm, combined with time series model TS(t) and neural network model to predict future blood sugar change trend:

[0047]

[0048] in is the predicted blood glucose concentration, ∈ is the error, and the neural network model is a hybrid neural network combined with a long short-term memory network, which is used to capture the correlation between time series and static features. In order to improve the real-time prediction of future blood glucose change trends, the dynamic trend analysis algorithm adopts a feature selection method based on causal inference, and dynamically adjusts the model weights in combination with the user's current physiological state and external environmental parameters, so that it can respond quickly in emergencies and provide higher prediction accuracy.

[0049] h. Determine whether the current and predicted blood sugar status is within the personalized blood sugar safety range [G min ,G max ], generate preliminary warning signal W, personalized blood sugar safety range [G min ,G max ] can be dynamically optimized based on the user’s physical condition, age, and daily activity patterns. The Bayesian decision model input is combined with the historical warning false alarm rate to reduce the frequency of false alarms and improve the credibility of warnings.

[0050] i. Combined with the user's daily activity records, dietary intake data and physiological cycle parameters, the accuracy probability of the warning is calculated through the Bayesian decision model P(W|A,C). The calculation of the accuracy probability of the warning also comprehensively considers the user's historical false alarm feedback data, and dynamically updates the parameter distribution of the Bayesian decision model through the reinforcement learning framework to further reduce the false alarm rate and improve the model's adaptability to complex physiological conditions.

[0051] j. Determine whether to send a warning signal and its level based on the probability P(W) and the warning level threshold θ;

[0052] k. Send personalized blood sugar management suggestions M, including diet, exercise and medication adjustment plans, via vibration, sound or mobile phone notification according to the warning level:

[0053] M=f(W,user profile,activitylog);

[0054] The warning signal is presented in three ways: visual charts, voice prompts and emergency vibration reminders. Users can set the priority by themselves. In step k, the management suggestion M is generated based on user feedback data and doctor's suggestions, and can be sent to the medical service platform through remote synchronization. The content of the management suggestion can be dynamically adjusted based on real-time data, and through remote interaction with doctors or medical platforms, the distribution information of medical resources in the user's area can be added to provide a quick solution for medical intervention in emergency situations.

[0055] I. Users can view blood sugar status, trend analysis and intervention suggestions through the mobile interface, and can upload feedback information and update model parameters. The early warning method performs real-time data processing through the edge computing platform and performs periodic big data analysis through the cloud server to improve algorithm stability. The early warning method is compatible with wearable devices, including smart bracelets, smart watches and independent blood sugar testers, and supports linkage with other health monitoring devices. The early warning method adopts a dynamic parameter tuning mechanism, regularly evaluates the interference compensation effect, and automatically adjusts the collection and calculation parameters to adapt to environmental changes. The early warning method supports the generation of personalized health reports based on artificial intelligence, and provides users with weekly or monthly blood sugar change summaries and health recommendations.

[0056] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A blood sugar monitoring and early warning method, characterized in that: The following steps are involved: a. Collect the skin surface spectral signal I(λ), where λ is the spectral wavelength, combining the near-infrared spectrum and the visible light spectrum signal; b. Collect skin surface temperature T, humidity H, ambient light intensity L and ambient electromagnetic interference E parameters; c. Preprocess the spectral signal I(λ), including removing noise N(I) through wavelet transform and correcting the signal intensity through the ambient light intensity compensation model to obtain the purified spectral signal: I′(λ)=I(λ)-N(I)+L c , where L c is the light intensity compensation coefficient; d. Based on the T, H, L and E parameters, the comprehensive interference compensation coefficient C is calculated using the multimodal regression model f(T,H,L,E) f ; e. The interference compensation coefficient C f Introduce the spectral signal correction formula: I c (λ)=I′(λ)×(1-C f ); Get the corrected spectral signal I c (λ); f. From I c (λ) extracts the spectral feature vector X = {x1, x2, …, x n }, combined with the user's historical blood sugar data vector Y = {y1,y2,…,y n } and individual biological parameters for feature matching; g. Use dynamic trend analysis algorithm, combined with time series model TS(t) and neural network model to predict future blood sugar change trend: in is the predicted blood glucose concentration, ∈ is the error; h. Determine whether the current and predicted blood sugar status is within the personalized blood sugar safety range [G min ,G max ], generating a preliminary warning signal W; i. Combine the user's daily activity records, dietary intake data and physiological cycle parameters to calculate the accuracy probability of the warning through the Bayesian decision model P(W|A,C); j. Determine whether to send a warning signal and its level based on the probability P(W) and the warning level threshold θ; k. Send personalized blood sugar management suggestions M, including diet, exercise and medication adjustment plans, via vibration, sound or mobile phone notification according to the warning level: M=f(W,user profile,activitylog); I. Users can view blood sugar status, trend analysis and intervention suggestions through the mobile interface, and can upload feedback information and update model parameters.

2. A blood sugar monitoring and early warning method according to claim 1, characterized in that: The spectral signal I(λ) collected in step a uses a dual-wavelength alternating sampling technique, combined with a fixed wavelength and an adjustable wavelength signal enhancement mode, to improve the sensitivity of the spectral data. The compensation of the ambient light intensity L in step b is achieved by adaptively adjusting the model L c =g(L), which can dynamically adapt to changes in indoor and outdoor collection environments.

3. A blood sugar monitoring and early warning method according to claim 1, characterized in that: The spectral signal purification in step c combines empirical mode decomposition and fast Fourier transform to optimize the denoising process, thereby improving the timeliness and stability of the purified signal. The comprehensive interference compensation coefficient C in step d f It is trained using a convolutional neural network, and the model input includes temperature, humidity, light intensity and electromagnetic interference parameters.

4. A blood sugar monitoring and early warning method according to claim 1, characterized in that: The individual biological parameters include body mass index, insulin sensitivity and long-term blood sugar control index. The neural network model in step g is a hybrid neural network combined with a long short-term memory network, which is used to capture the correlation between time series and static features.

5. A blood sugar monitoring and early warning method according to claim 1, characterized in that: The personalized blood glucose safety range described in step h [G min ,G max ] can be dynamically optimized based on the user's physique, age and daily activity patterns. The Bayesian decision model input is combined with the historical warning false alarm rate to reduce the frequency of false alarms and improve the credibility of warnings.

6. A blood sugar monitoring and early warning method according to claim 1, characterized in that: The warning signal is presented in three ways: visual chart, voice prompt and emergency vibration reminder. The user can set the priority by himself. The management suggestion M in step k is generated by combining user feedback data and doctor's advice, and can be sent to the medical service platform through remote synchronization.

7. A blood sugar monitoring and early warning method according to claim 1, characterized in that: The early warning method performs real-time data processing through an edge computing platform, and performs periodic big data analysis through a cloud server to improve algorithm stability. The early warning method is suitable for wearable devices, including smart bracelets, smart watches and independent blood glucose monitors, and supports linkage with other health monitoring devices.

8. A blood sugar monitoring and early warning method according to claim 1, characterized in that: The early warning method adopts a dynamic parameter tuning mechanism, regularly evaluates the interference compensation effect, and automatically adjusts the acquisition and calculation parameters to adapt to environmental changes. The early warning method supports the generation of personalized health reports based on artificial intelligence, and provides users with weekly or monthly blood sugar change summaries and health recommendations.

Images