Blood glucose detection method and blood glucose detector for non-invasive blood glucose meter

Abstract

The present application relates to non-invasive blood glucose detection technology, and particularly relates to a blood glucose detection method and a blood glucose detector for a non-invasive blood glucose meter. The blood glucose detection method for the non-invasive blood glucose meter comprises the following steps: collecting near-infrared light data and mid-infrared light data and calculating corresponding weight coefficients; obtaining a predicted blood glucose value by using a blood glucose detection model, calculating a credibility, comparing the credibility with a preset credibility threshold, and performing corresponding operations; and obtaining a blood glucose health assessment result by using a blood glucose health assessment model. The present application combines near-infrared spectroscopy technology and mid-infrared spectroscopy technology, and considers the influence of environmental factors in real time, calculates the weight coefficients of the near-infrared light data and the mid-infrared light data respectively, adjusts and optimizes the analysis of the spectral data, provides more accurate blood glucose detection results, and ensures that reliable blood glucose measurement values can be obtained under different environmental conditions.

Description

Technical Field

[0001] This invention relates to non-invasive blood glucose testing technology, specifically to a blood glucose testing method and a blood glucose meter for use in a non-invasive blood glucose meter. Background Technology

[0002] The incidence of metabolic diseases such as diabetes is rising year by year, and blood glucose monitoring has become an indispensable part of the daily management of diabetic patients. Traditional blood glucose monitoring methods mainly rely on invasive finger prick blood sampling, which not only causes pain and inconvenience to patients, but may also lead to the risk of infection. In addition, for patients with brain diseases, such as stroke or brain injury patients, they need to regularly test their blood glucose at home to ensure safety and adjust their treatment plan in a timely manner. These patients may have difficulty moving and are more sensitive to pain, so traditional invasive blood glucose monitoring methods are particularly inconvenient for them.

[0003] To address this issue, developing a non-invasive, accurate, and convenient blood glucose testing technology has become an urgent need in the medical and health field. Currently, various non-invasive blood glucose testing technologies are available on the market, such as optical, ultrasonic, and electromagnetic methods. Among them, optical methods have received widespread attention due to their non-invasiveness, rapid response, and high sensitivity. Optical methods mainly infer blood glucose concentration by analyzing the propagation characteristics of light in biological tissues, such as absorption and scattering. However, existing optical non-invasive blood glucose testing technologies still face some challenges, including the influence of environmental factors on measurement results and how to improve the accuracy and reliability of measurements. To overcome these challenges, this invention aims to provide a more advanced, accurate, and user-friendly non-invasive blood glucose testing method to meet the daily blood glucose monitoring needs of diabetic patients, especially those with limited mobility due to brain diseases, and to bring innovation to the medical and health field. Summary of the Invention

[0004] This invention combines near-infrared spectroscopy and mid-infrared spectroscopy, and takes into account the influence of environmental factors such as temperature, humidity and light intensity on the measurement results in real time. It calculates the weighting coefficients of near-infrared light data and mid-infrared light data respectively, thereby adjusting and optimizing the analysis of spectral data, providing more accurate blood glucose detection results, and ensuring that reliable blood glucose measurement values ​​can be obtained under different environmental conditions, making blood glucose prediction and health assessment more accurate.

[0005] A blood glucose detection method for use in a non-invasive blood glucose meter, comprising:

[0006] When a user uses a blood glucose meter to test blood glucose, the meter collects near-infrared light data and mid-infrared light data using near-infrared spectroscopy and mid-infrared spectroscopy, respectively. At the same time, it acquires temperature data, humidity data, and light interference intensity data of the current collection environment. The acquired temperature data, humidity data, and light interference intensity data are used as environmental interference data. The weighting coefficients of near-infrared light data and mid-infrared light data are calculated using the environmental interference data.

[0007] Near-infrared light data, mid-infrared light data, and corresponding weighting coefficients are input into the blood glucose detection model to output a predicted blood glucose value. At the same time, the reliability of the current predicted blood glucose value is calculated using the acquired environmental interference data. It is then determined whether the reliability of the current predicted blood glucose value is greater than a preset reliability threshold. If so, the predicted blood glucose value is displayed on the blood glucose meter. If not, the predicted blood glucose value and reliability are displayed on the blood glucose meter simultaneously, and the user is prompted that the reliability of the current predicted blood glucose value is low and it is recommended to change the detection environment.

[0008] The acquired predicted blood glucose values ​​and their corresponding confidence levels are transmitted to the cloud server. At the same time, the detection time of the predicted blood glucose values, the user's most recent eating time, and the user's exercise status before the predicted blood glucose values ​​are recorded. Several recently acquired predicted blood glucose values, their corresponding confidence levels, detection times, most recent eating times, and exercise status are arranged in chronological order to form a blood glucose health assessment dataset. The blood glucose health assessment dataset is used as the input to the blood glucose health assessment model, and the blood glucose health assessment results are output.

[0009] Preferably, the weighting coefficients for near-infrared light data and mid-infrared light data are calculated using environmental interference data, as follows:

[0010] Based on the environmental interference data, including temperature data T, humidity data H, and light interference intensity data L; for temperature data T, a suitable temperature range is set as follows: to If the temperature data T is located at to Between, the temperature effect index The value is 0; if the temperature data T is less than or equal to Then use the formula Calculate and obtain the temperature influence index If the temperature data T is greater than or equal to Then use the formula Calculate and obtain the temperature influence index ;

[0011] For humidity data H, set the suitable humidity range as follows: to If the humidity data H is located in to Between, the humidity effect index The value is 0; if the humidity data H is less than or equal to Then use the formula Calculate and obtain the humidity impact index If the humidity data H is greater than or equal to Then use the formula Calculate and obtain the humidity impact index ;

[0012] Set a standard light intensity for the light interference intensity data L. If the light interference intensity data L is less than or equal to the standard light intensity The light intensity influence index The value is 0; if the light interference intensity data L is greater than the standard light intensity... Then use the formula Calculate and obtain the light intensity influence index ;

[0013] The relative contribution coefficients of temperature, humidity, and light interference intensity to near-infrared light data were set as follows: , and Meanwhile, the relative contribution coefficients of temperature, humidity, and light interference intensity to mid-infrared light data were set to [value missing]. , and ,in, < , < , > Using formulas Calculate the impact index on the acquisition of near-infrared light data Using the formula Calculate the impact index on mid-infrared light data acquisition Finally, use the formula Calculate the weighting coefficients for acquiring near-infrared light data Weighting coefficients of mid-infrared light data Through formula Calculated and obtained.

[0014] Preferably, the blood glucose detection model is based on a fully connected neural network, including a first input layer, a first fully connected layer, a feature weighting layer, a second fully connected layer, and a first output layer; the first input layer is used to input near-infrared light data, mid-infrared light data, and corresponding weight coefficients; the first fully connected layer is used to extract features from the near-infrared light data and mid-infrared light data; the feature weighting layer is used to apply the calculated weight coefficients to weight the features of the near-infrared light data and mid-infrared light data; the second fully connected layer is used to further extract features from the weighted features of the near-infrared light data and mid-infrared light data; and the first output layer is used to output the predicted blood glucose value.

[0015] Preferably, the specific steps for training the blood glucose detection model are as follows:

[0016] Obtain several labeled blood glucose test training samples. The labels represent the actual blood glucose values ​​corresponding to the training samples. Each training sample contains a set of near-infrared light data, mid-infrared light data, and weight coefficients for the infrared and mid-infrared light data. Divide all blood glucose test training samples into a blood glucose test training set and a blood glucose test validation set. Train the blood glucose test model using the training set and validate the model using the validation set to obtain the first validation result. Determine whether the first validation result meets the preset first training condition. If it does, output the trained blood glucose test model; otherwise, continue training the blood glucose test model using the training set.

[0017] Preferably, the confidence level of the current predicted blood glucose value is calculated using the acquired environmental interference data, and the specific operation is as follows:

[0018] Regarding temperature data T, within a suitable temperature range of... to Based on this, the temperature range is set as follows: to When the temperature data T is within the suitable temperature range to Temperature rating during the period The value is 1; when the temperature data T exceeds the temperature range. to Temperature rating The value is 0; when the temperature data T is located at to When, use the formula Calculate and obtain temperature score When the temperature data T is located at to When in between, use the formula Calculate and obtain temperature score ;

[0019] Regarding humidity data H, within the suitable humidity range is to Based on this, the humidity usage range is set as follows: to When the humidity data H is within the suitable humidity range to Humidity score The value is 1; when the humidity data H exceeds the humidity usage range. to At that time, humidity score The value is 0; when the humidity data H is located at to When, use the formula Calculate and obtain humidity score When humidity data H is located to When in between, use the formula Calculate and obtain humidity score ;

[0020] Regarding the light interference intensity data L, under standard light intensity Based on this, set limits on light intensity. When the light interference intensity data L is less than or equal to the standard light intensity At that time, light intensity score The value is 1; when the light interference intensity data L is at the standard light intensity With limiting light intensity When in between, use the formula Calculate and obtain light intensity score When the light interference intensity data L is greater than or equal to the limiting light intensity At that time, light intensity score The value is 0;

[0021] Finally, use the formula Calculate the confidence level S of the currently obtained predicted blood glucose value; where , and The weights for the effects of temperature, humidity, and light interference intensity on the accuracy of blood glucose prediction are respectively: + + =1.

[0022] Preferably, the blood glucose health assessment model is based on an LSTM model, including a second input layer, an LSTM layer, a third fully connected layer, and a second output layer; the second input layer is used to receive the input blood glucose health assessment dataset; the LSTM layer is used to extract the temporal features of the blood glucose health assessment dataset; the third fully connected layer is used to further extract features from the output of the LSTM layer; and the second output layer is used to output the blood glucose health assessment results.

[0023] Preferably, the specific steps for training the blood glucose health assessment model are as follows:

[0024] Obtain several labeled blood glucose health assessment training samples. Each blood glucose health assessment training sample contains a set of predicted blood glucose values ​​sorted by time, along with the corresponding confidence level, detection time, recent eating time, and exercise status. Divide all health assessment training samples into a health assessment training set and a health assessment validation set. Use the predicted blood glucose values ​​sorted by time in the health assessment training set, along with the corresponding confidence level, detection time, recent eating time, and exercise status, as input to the blood glucose health assessment model. Use the blood glucose health assessment results as the output of the blood glucose health assessment model to train the model. Validate the blood glucose health assessment model using the health assessment validation set to obtain a second validation result. Determine whether the second validation result meets the preset second training conditions. If yes, output the trained blood glucose health assessment model; otherwise, continue training the blood glucose health assessment model using the health assessment training set.

[0025] Preferably, a blood glucose meter is provided, wherein the blood glucose meter is applied to any of the above-mentioned blood glucose testing methods for non-invasive blood glucose meters.

[0026] The present invention has the following advantages:

[0027] 1. This invention combines near-infrared spectroscopy and mid-infrared spectroscopy, and takes into account the influence of environmental factors such as temperature, humidity and light intensity on the measurement results in real time. It calculates the weighting coefficients of near-infrared light data and mid-infrared light data respectively, thereby adjusting and optimizing the analysis of spectral data, providing more accurate blood glucose detection results, and ensuring that reliable blood glucose measurement values ​​can be obtained under different environmental conditions, making blood glucose prediction and health assessment more accurate.

[0028] 2. This invention records and analyzes users' blood glucose data through a cloud server, analyzes it using a blood glucose health assessment model, obtains blood glucose health assessment results, and provides users with personalized health management suggestions, enhancing users' understanding and management capabilities of their own health status. This convenient and user-friendly design makes this invention very suitable for daily home use, improving patients' quality of life and the efficiency of diabetes management. Attached Figure Description

[0029] Figure 1 This is a schematic flowchart of the blood glucose detection method for a non-invasive blood glucose meter used in an embodiment of the present invention.

[0030] Figure 2 This is a rear view of the blood glucose meter of the present invention;

[0031] Figure 3 This is a bottom view of the blood glucose meter of the present invention;

[0032] Figure 4 This is a top view of the blood glucose meter of the present invention;

[0033] Figure 5 This is a first perspective view of the blood glucose meter of the present invention;

[0034] Figure 6 This is a second perspective view of the blood glucose meter of the present invention;

[0035] Figure 7 This is a right view of the blood glucose meter of the present invention;

[0036] Figure 8 This is a forward view of the blood glucose meter of the present invention;

[0037] Figure 9 This is a left view of the blood glucose meter of the present invention. Detailed Implementation

[0038] To enable those skilled in the art to better understand the technical solutions of this invention, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of this invention.

[0039] Example: A blood glucose detection method for a non-invasive blood glucose meter, such as... Figure 1 As shown, it includes:

[0040] For patients with brain diseases, such as stroke or brain injury, regular blood glucose monitoring at home is necessary to ensure safety and allow for timely adjustments to treatment plans. These patients may have limited mobility and are more sensitive to pain; therefore, the blood glucose testing method for non-invasive blood glucose meters of this invention can be applied. When the user uses the blood glucose meter to test blood glucose, the meter collects near-infrared and mid-infrared light data using near-infrared and mid-infrared spectroscopy technologies, respectively. Near-infrared light can penetrate the skin surface and interact with glucose molecules in subcutaneous tissue, thus carrying information about blood glucose levels. Mid-infrared light is more sensitive to the absorption of glucose molecules, thus providing more direct... The system acquires blood glucose concentration information; simultaneously, it obtains temperature, humidity, and light interference intensity data of the current collection environment. These data are used as environmental interference data, as ambient temperature affects light propagation in the skin and the activity of glucose molecules, making temperature data crucial for calibrating measurement results. Skin surface humidity affects light scattering and absorption, thus impacting the accuracy of blood glucose measurements. External light source interference may affect the acquisition of spectral data, therefore, light interference intensity needs to be measured and considered. Weighting coefficients for near-infrared and mid-infrared light data are calculated using the environmental interference data to reduce the impact of environmental factors on blood glucose measurement results.

[0041] Near-infrared and mid-infrared light data, along with corresponding weighting coefficients, are input into a blood glucose detection model to output a predicted blood glucose value. Simultaneously, using acquired environmental interference data, the reliability of the current predicted blood glucose value is calculated. Reliability is a quantitative indicator reflecting the reliability of blood glucose measurement results under current environmental conditions. Furthermore, the reliability of the predicted blood glucose value plays a crucial role in subsequent blood glucose health assessment model analysis. It not only affects the judgment of blood glucose status but also helps the model more accurately understand the reliability of different measurement data. Blood glucose values ​​with higher reliability have a greater impact on the blood glucose health assessment model, while those with lower reliability have a smaller impact on the results. This helps avoid the influence of low-reliability data on blood glucose health. The system mitigates interference during the training and evaluation of the blood glucose assessment model, thereby better assessing and predicting blood glucose health status. It determines whether the reliability of the current predicted blood glucose value exceeds a preset reliability threshold. If so, the predicted blood glucose value is displayed on the blood glucose meter; otherwise, both the predicted blood glucose value and the reliability are displayed on the meter, and the user is alerted that the current predicted blood glucose value has low reliability, suggesting a change of testing environment. In this way, the blood glucose meter not only provides predicted blood glucose values ​​but also assesses the reliability of measurement results, ensuring users receive accurate and reliable blood glucose monitoring information. This design improves the practicality of non-invasive blood glucose testing and user trust, while also providing users with suggestions for environmental adjustments to obtain better measurement results.

[0042] The acquired predicted blood glucose values ​​and their corresponding confidence levels are transmitted to a cloud server for further analysis and storage. The cloud server provides a centralized platform for collecting and processing data from multiple users, while recording the detection time of the predicted blood glucose values, the user's most recent meal time, and the user's exercise status before obtaining the predicted blood glucose values. This information is recorded along with the predicted blood glucose values ​​because it may affect blood glucose levels; for example, blood glucose levels usually rise after eating and may fall after exercise. Several recently acquired predicted blood glucose values, along with their corresponding confidence levels, detection times, most recent meal times, and exercise status, are arranged chronologically to form a blood glucose health assessment dataset. This dataset provides a comprehensive perspective for analyzing the user's blood glucose change trends and patterns. The blood glucose health assessment dataset is used as input to the blood glucose health assessment model, which outputs a blood glucose health assessment result. This result is an overall evaluation of blood glucose control, helping users understand their own blood glucose management and adjust their treatment plans in a timely manner.

[0043] The weighting coefficients for near-infrared and mid-infrared light data are calculated using environmental interference data, as follows:

[0044] Based on the environmental interference data, including temperature data T, humidity data H, and light interference intensity data L; for temperature data T, a suitable temperature range is set as follows: to If the temperature data T is located at to Between, the temperature effect index The value is 0; if the temperature data T is less than or equal to Then use the formula Calculate and obtain the temperature influence index If the temperature data T is greater than or equal to Then use the formula Calculate and obtain the temperature influence index ;

[0045] For humidity data H, set the suitable humidity range as follows: to If the humidity data H is located at to Between, the humidity effect index The value is 0; if the humidity data H is less than or equal to Then use the formula Calculate and obtain the humidity impact index If the humidity data H is greater than or equal to Then use the formula Calculate and obtain the humidity impact index ;

[0046] Set a standard light intensity for the light interference intensity data L. If the light interference intensity data L is less than or equal to the standard light intensity The light intensity influence index The value is 0; if the light interference intensity data L is greater than the standard light intensity... Then use the formula Calculate and obtain the light intensity influence index ;

[0047] The relative contribution coefficients of temperature, humidity, and light interference intensity to near-infrared light data were set as follows: , and Meanwhile, the relative contribution coefficients of temperature, humidity, and light interference intensity to mid-infrared light data were set to [value missing]. , and Where NIR represents near-infrared and MIR represents mid-infrared. < , < , > Because temperature affects the absorption and scattering characteristics of infrared light; mid-infrared light is generally more sensitive to temperature changes, so the weight of mid-infrared light data may need to be reduced under high or unstable temperature conditions; changes in humidity may affect both near-infrared and mid-infrared spectra, but the effect on mid-infrared light is usually more significant because the absorption peak of moisture appears in the mid-infrared region; changes in ambient light may have a greater impact on near-infrared light because near-infrared light has a shorter wavelength and is easily interfered with by visible light.

[0048] Using formula Calculate the impact index on the acquisition of near-infrared light data Using the formula Calculate the impact index on mid-infrared light data acquisition Finally, use the formula Calculate the weighting coefficients for acquiring near-infrared light data Weighting coefficients of mid-infrared light data Through formula Calculated and obtained.

[0049] The blood glucose detection model is based on a fully connected neural network, a fundamental deep learning model composed of multiple fully connected layers. Each neuron in this network is connected to all neurons in the previous layer, forming a dense network structure. This structure allows information to flow freely within the network, unrestricted by spatial relationships, enabling it to learn complex nonlinear relationships and patterns. Fully connected neural networks are commonly used to process structured data, such as tabular data, and can also be used as part of other complex network structures for feature extraction and classification tasks. During training, the prediction results are calculated through forward propagation, and then the network weights are adjusted through backpropagation. To minimize prediction error and thus improve the model's accuracy and generalization ability, the blood glucose detection model includes a first input layer, a first fully connected layer, a feature weighting layer, a second fully connected layer, and a first output layer. The first input layer is used to input near-infrared light data, mid-infrared light data, and corresponding weight coefficients. The first fully connected layer is used to extract features from the near-infrared and mid-infrared light data. The feature weighting layer is used to apply the calculated weight coefficients to weight the features of the near-infrared and mid-infrared light data. The second fully connected layer is used to further extract features from the weighted near-infrared and mid-infrared light data. The first output layer is used to output the predicted blood glucose value.

[0050] The specific steps for training the blood glucose detection model are as follows:

[0051] Obtain several labeled blood glucose test training samples. The labels represent the actual blood glucose values ​​corresponding to the training samples. Each training sample contains a set of near-infrared light data, mid-infrared light data, and weight coefficients for the infrared and mid-infrared light data. The pre-calculated weight coefficients remain unchanged during training and are only used to weight the features. Divide all blood glucose test training samples into a blood glucose test training set and a blood glucose test validation set. Train the blood glucose test model using the training set and validate the model using the validation set to obtain the first validation result. Determine whether the first validation result meets the preset first training condition. If it does, output the trained blood glucose test model; otherwise, continue training the blood glucose test model using the training set.

[0052] Using the acquired environmental interference data, the confidence level of the current predicted blood glucose value is calculated. The specific steps are as follows:

[0053] Regarding temperature data T, within a suitable temperature range of... to Based on this, the temperature range is set as follows: to When the temperature data T is within the suitable temperature range to Temperature rating during the period The value is 1; when the temperature data T exceeds the temperature range. to Temperature rating The value is 0; when the temperature data T is located at to When, use the formula Calculate and obtain temperature score When the temperature data T is located at to When in between, use the formula Calculate and obtain temperature score ;

[0054] Regarding humidity data H, within the suitable humidity range is to Based on this, the humidity usage range is set as follows: to When the humidity data H is within the suitable humidity range to Humidity score The value is 1; when the humidity data H exceeds the humidity usage range. to At that time, humidity score The value is 0; when the humidity data H is located at to When, use the formula Calculate and obtain humidity score When humidity data H is located to When in between, use the formula Calculate and obtain humidity score ;

[0055] Regarding the light interference intensity data L, under standard light intensity Based on this, set limits on light intensity. When the light interference intensity data L is less than or equal to the standard light intensity At that time, light intensity score The value is 1; when the light interference intensity data L is at the standard light intensity With limiting light intensity When in between, use the formula Calculate and obtain light intensity score When the light interference intensity data L is greater than or equal to the limiting light intensity At that time, light intensity score The value is 0;

[0056] Finally, use the formula Calculate the confidence level S of the currently obtained predicted blood glucose value; where , and The weights for the effects of temperature, humidity, and light interference intensity on the accuracy of blood glucose prediction are respectively: + + =1.

[0057] The blood glucose health assessment model is based on the LSTM model, a special type of recurrent neural network capable of learning and memorizing long-term dependent information. LSTM controls the flow of information by introducing input gates, forget gates, and output gates. These gating mechanisms allow the network to selectively retain or discard information, effectively solving the gradient vanishing and gradient exploding problems of traditional RNNs when processing long sequence data. The blood glucose health assessment model includes a second input layer, an LSTM layer, a third fully connected layer, and a second output layer. The second input layer receives the input blood glucose health assessment dataset; the LSTM layer extracts the temporal features of the blood glucose health assessment dataset; the third fully connected layer further extracts features from the output of the LSTM layer; and the second output layer outputs the blood glucose health assessment results.

[0058] The specific steps for training the blood glucose health assessment model are as follows:

[0059] Obtain several labeled blood glucose health assessment training samples. Each blood glucose health assessment training sample contains a set of predicted blood glucose values ​​sorted by time, along with the corresponding confidence level, detection time, recent eating time, and exercise status. Divide all health assessment training samples into a health assessment training set and a health assessment validation set. Use the predicted blood glucose values ​​sorted by time in the health assessment training set, along with the corresponding confidence level, detection time, recent eating time, and exercise status, as input to the blood glucose health assessment model. Use the blood glucose health assessment results as the output of the blood glucose health assessment model to train the model. Validate the blood glucose health assessment model using the health assessment validation set to obtain a second validation result. Determine whether the second validation result meets the preset second training conditions. If yes, output the trained blood glucose health assessment model; otherwise, continue training the blood glucose health assessment model using the health assessment training set.

[0060] A blood glucose meter, such as Figures 2-9 As shown, the blood glucose meter is applied to any of the above-mentioned blood glucose testing methods for non-invasive blood glucose meters.

[0061] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims. Parts not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method for blood glucose detection using a non-invasive blood glucose meter, characterized in that, include: When a user uses a blood glucose meter to test blood glucose, the meter collects near-infrared light data and mid-infrared light data using near-infrared spectroscopy and mid-infrared spectroscopy, respectively. At the same time, it acquires temperature data, humidity data, and light interference intensity data of the current collection environment. The acquired temperature data, humidity data, and light interference intensity data are used as environmental interference data. The weighting coefficients of near-infrared light data and mid-infrared light data are calculated using the environmental interference data. Near-infrared light data, mid-infrared light data, and corresponding weighting coefficients are input into the blood glucose detection model to output a predicted blood glucose value. At the same time, the reliability of the current predicted blood glucose value is calculated using the acquired environmental interference data. It is then determined whether the reliability of the current predicted blood glucose value is greater than a preset reliability threshold. If so, the predicted blood glucose value is displayed on the blood glucose meter. If not, the predicted blood glucose value and reliability are displayed on the blood glucose meter simultaneously, and the user is prompted that the reliability of the current predicted blood glucose value is low and it is recommended to change the detection environment. The acquired predicted blood glucose values ​​and their corresponding confidence levels are transmitted to the cloud server. At the same time, the detection time of the predicted blood glucose values, the user's most recent eating time, and the user's exercise status before the predicted blood glucose values ​​are recorded. Several recently acquired predicted blood glucose values, their corresponding confidence levels, detection times, most recent eating times, and exercise status are arranged in chronological order to form a blood glucose health assessment dataset. The blood glucose health assessment dataset is used as the input to the blood glucose health assessment model, and the blood glucose health assessment results are output.

2. The blood glucose detection method for a non-invasive blood glucose meter according to claim 1, characterized in that, The weighting coefficients for near-infrared and mid-infrared light data are calculated using environmental interference data, as follows: Based on the environmental interference data, including temperature data T, humidity data H, and light interference intensity data L; for temperature data T, a suitable temperature range is set as follows: to If the temperature data T is located at to Between, the temperature effect index The value is 0; if the temperature data T is less than or equal to Then use the formula Calculate and obtain the temperature influence index If the temperature data T is greater than or equal to Then use the formula Calculate and obtain the temperature influence index ; For humidity data H, set the suitable humidity range as follows: to If the humidity data H is located in to Between, the humidity impact index The value is 0; if the humidity data H is less than or equal to Then use the formula Calculate and obtain the humidity impact index If the humidity data H is greater than or equal to Then use the formula Calculate and obtain the humidity impact index ; Set a standard light intensity for the light interference intensity data L. If the light interference intensity data L is less than or equal to the standard light intensity The light intensity influence index The value is 0; if the light interference intensity data L is greater than the standard light intensity... Then use the formula Calculate and obtain the light intensity influence index ; The relative contribution coefficients of temperature, humidity, and light interference intensity to near-infrared light data were set as follows: , and Meanwhile, the relative contribution coefficients of temperature, humidity, and light interference intensity to mid-infrared light data were set to [value missing]. , and ,in, < , < , > Using formulas Calculate the impact index on the acquisition of near-infrared light data Using the formula Calculate the impact index on mid-infrared light data acquisition Finally, use the formula Calculate the weighting coefficients for acquiring near-infrared light data Weighting coefficients of mid-infrared light data Through formula Calculated and obtained.

3. The blood glucose detection method for a non-invasive blood glucose meter according to claim 2, characterized in that, The blood glucose detection model is based on a fully connected neural network, including a first input layer, a first fully connected layer, a feature weighting layer, a second fully connected layer, and a first output layer. The first input layer is used to input near-infrared light data, mid-infrared light data, and corresponding weight coefficients. The first fully connected layer is used to extract features from the near-infrared light data and mid-infrared light data. The feature weighting layer is used to apply the calculated weight coefficients to weight the near-infrared light data features and mid-infrared light data features; the second fully connected layer is used to further extract features from the weighted near-infrared light data features and mid-infrared light data features; the first output layer is used to output the predicted blood glucose value.

4. A blood glucose detection method for a non-invasive blood glucose meter according to claim 3, characterized in that, The specific steps for training the blood glucose detection model are as follows: Obtain several labeled blood glucose test training samples. The labels represent the actual blood glucose values ​​corresponding to the training samples. Each training sample contains a set of near-infrared light data, mid-infrared light data, and weight coefficients for the infrared and mid-infrared light data. Divide all blood glucose test training samples into a blood glucose test training set and a blood glucose test validation set. Train the blood glucose test model using the training set and validate the model using the validation set to obtain the first validation result. Determine whether the first validation result meets the preset first training condition. If it does, output the trained blood glucose test model; otherwise, continue training the blood glucose test model using the training set.

5. A blood glucose detection method for a non-invasive blood glucose meter according to claim 4, characterized in that, Using the acquired environmental interference data, the confidence level of the current predicted blood glucose value is calculated. The specific steps are as follows: Regarding temperature data T, within a suitable temperature range of... to Based on this, the temperature range is set as follows: to When the temperature data T is within the suitable temperature range to Temperature rating during the period The value is 1; when the temperature data T exceeds the temperature range. to Temperature rating The value is 0; when the temperature data T is located at to When, use the formula Calculate and obtain temperature score When the temperature data T is located at to When in between, use the formula Calculate and obtain temperature score ; Regarding humidity data H, within the suitable humidity range is to Based on this, the humidity usage range is set as follows: to When the humidity data H is within the suitable humidity range to Humidity score The value is 1; when the humidity data H exceeds the humidity usage range. to At that time, humidity score The value is 0; when the humidity data H is located at to When, use the formula Calculate and obtain humidity score When humidity data H is located to When in between, use the formula Calculate and obtain humidity score ; Regarding the light interference intensity data L, under standard light intensity Based on this, set limits on light intensity. ; When the light interference intensity data L is less than or equal to the standard light intensity Light intensity score The value is 1; when the light interference intensity data L is at the standard light intensity With limiting light intensity When in between, use the formula Calculate and obtain light intensity score ; When the light interference intensity data L is greater than or equal to the light intensity limit Light intensity score The value is 0; Finally, use the formula Calculate the confidence level S of the currently obtained predicted blood glucose value; where , and The weights for the effects of temperature, humidity, and light interference intensity on the accuracy of blood glucose prediction are respectively: + + =1.

6. A blood glucose detection method for a non-invasive blood glucose meter according to claim 5, characterized in that, The blood glucose health assessment model is based on an LSTM model, consisting of a second input layer, an LSTM layer, a third fully connected layer, and a second output layer. The second input layer receives the blood glucose health assessment dataset as input, and the LSTM layer extracts the temporal features of the blood glucose health assessment dataset. The third fully connected layer is used to further extract features from the output of the LSTM layer; The second output layer is used to output blood glucose health assessment results.

7. A blood glucose detection method for a non-invasive blood glucose meter according to claim 6, characterized in that, The specific steps for training the blood glucose health assessment model are as follows: Obtain several labeled blood glucose health assessment training samples. Each blood glucose health assessment training sample contains a set of predicted blood glucose values ​​sorted by time, along with the corresponding confidence level, detection time, recent eating time, and exercise status. Divide all health assessment training samples into a health assessment training set and a health assessment validation set. Use the predicted blood glucose values ​​sorted by time in the health assessment training set, along with the corresponding confidence level, detection time, recent eating time, and exercise status, as input to the blood glucose health assessment model. Use the blood glucose health assessment results as the output of the blood glucose health assessment model to train the model. Validate the blood glucose health assessment model using the health assessment validation set to obtain a second validation result. Determine whether the second validation result meets the preset second training conditions. If yes, output the trained blood glucose health assessment model; otherwise, continue training the blood glucose health assessment model using the health assessment training set.

8. A blood glucose meter, characterized in that, The blood glucose meter is used in the blood glucose testing method for non-invasive blood glucose meters as described in any one of claims 1-7.

Images