Real-time intervention method and system based on blood glucose prediction of diabetic patient

Abstract

The invention provides a real-time intervention method and system based on blood glucose prediction of a diabetic patient, and the method comprises the steps: obtaining blood glucose monitoring data and multi-dimensional life monitoring data based on blood glucose monitoring equipment and intelligent equipment, and carrying out the timestamp alignment of the multi-dimensional life monitoring data and the blood glucose monitoring data; inputting the aligned multi-dimensional life monitoring data and blood glucose monitoring data into a pre-trained blood glucose prediction model, outputting a first blood glucose level prediction value and a second blood glucose level prediction value, and generating a real-time intervention suggestion according to the first blood glucose level prediction value and the second blood glucose level prediction value; if the first blood glucose level predicted value or the second blood glucose level predicted value is lower than a threshold value, graded early warning is carried out, corresponding intelligent equipment is triggered to execute corresponding response measures, and the early warning mechanism can achieve early warning through a real-time, normalized and highly-personalized intervention strategy in combination with intelligent operation of the intelligent equipment. The life quality of a diabetic patient is obviously improved; and the blood sugar control is optimized.

Description

Technical Field

[0001] The present invention relates to the field of digital medical technology, and in particular to a real-time intervention method and system based on blood sugar prediction for diabetic patients. Background Art

[0002] Diabetes is widespread around the world. According to statistics from the International Diabetes Federation (IDF Diabetes Atlas), the global prevalence of diabetes continues to rise. As of 2021, approximately 537 million adults aged 20 to 79 years old (10.5%) suffer from diabetes. The treatment and control of diabetes not only has an important impact on the quality of life of individual patients, but also puts tremendous pressure on the medical and social security systems.

[0003] The control of diabetes relies heavily on monitoring quantitative indicators (such as blood sugar) and self-management of diet, exercise, medication, etc. by patients, who have a high degree of autonomy. In the process of self-management, patients usually face the following problems: (1) Lack of professional medical knowledge, difficulty in interpreting changes in physiological indicators, and difficulty in adaptively adjusting behaviors such as diet, exercise, and medication; (2) Lack of sufficient positive feedback, incentives, and supervision, which makes it easy to be frustrated frequently in blood sugar management, thereby eroding enthusiasm and compliance with medical advice; (3) Lack of scientifically verified convenient, intelligent, and user-friendly blood sugar home management technology.

[0004] For blood sugar management, the main products currently available at home and abroad include Guardian Connect, RPM365, Sugar Nurse, Steady Sugar, Suitang, etc. These products lack real-time analysis of patients' blood sugar improvement behaviors during use, and thus cannot adaptively adjust blood sugar metabolism models, intervention strategies, and intervention methods. Summary of the invention

[0005] The present invention provides a real-time intervention method and system based on blood glucose prediction for diabetic patients, which is used to solve the technical defects in the prior art of lacking real-time and normalized analysis of patients' blood glucose improvement behavior during use, and being unable to adaptively adjust blood glucose metabolism models, intervention strategies, and intervention methods.

[0006] The present invention provides a real-time intervention method based on blood glucose prediction of diabetic patients, comprising: Based on the blood glucose monitoring device and the smart device, blood glucose monitoring data and multi-dimensional life monitoring data are obtained, and the multi-dimensional life monitoring data and the blood glucose monitoring data are time-stamped and aligned; Inputting the aligned multi-dimensional life monitoring data and blood glucose monitoring data into a pre-trained blood glucose prediction model, outputting a first blood glucose level prediction value and a second blood glucose level prediction value, wherein the prediction period of the first blood glucose level prediction value is shorter than the prediction period of the second blood glucose level prediction value; the blood glucose prediction model is trained based on blood glucose monitoring sample data and life monitoring sample data of diabetic patients; generating a real-time intervention recommendation based on the first blood glucose level prediction value and the second blood glucose level prediction value; If the first blood sugar level predicted value or the second blood sugar level predicted value is lower than a threshold, a graded warning is implemented, and the corresponding smart device is triggered to execute corresponding response measures.

[0007] According to the real-time intervention method provided by the present invention, the method further includes: when it is detected according to the blood sugar monitoring data and the multi-dimensional life monitoring data that the user is already in a hypoglycemic state, performing at least one of the following processing measures: Contact the user's family via phone and share the user's real-time location; Play voice messages through smart devices to alert people around to provide help; If it detects that there is no one around the user, it will automatically call the emergency number and explain the user's situation and location through the intelligent voice system embedded in the smart device.

[0008] According to the real-time intervention method provided by the present invention, if the first blood glucose level prediction value or the second blood glucose level prediction value is lower than the threshold, a graded warning is implemented, and the corresponding smart device is triggered to execute corresponding response measures, specifically including: If the second predicted value of the blood sugar level is lower than the threshold or the first predicted value of the blood sugar level is lower than the threshold, a graded warning is issued according to the distance between the user's current location and the residence: If the user's current location is where they live, a first-level warning will be issued through the smart device, and the user's family will be notified through push notifications; If the distance between the user's current location and the place of residence is less than or equal to the distance threshold, a second-level warning will be issued through the smart device, and the user's family will be informed through message push, and the user's current location will be shared in real time; If the distance between the user's current location and residence is greater than the distance threshold, a third-level warning will be issued through the smart device, and the user's family will be notified by phone, and the user's current location will be shared in real time.

[0009] According to the real-time intervention method provided by the present invention, after triggering the corresponding smart device to execute the corresponding response measures, the method also includes: triggering the smart home appliances connected to the smart device to execute the corresponding response measures; wherein the smart home appliances include at least one of the following: smart door locks, smart refrigerators, and smart cookers.

[0010] According to the real-time intervention method provided by the present invention, the multi-dimensional life monitoring data includes: dietary data, medication data and exercise data; after aligning the timestamps of the multi-dimensional life monitoring data with the blood glucose monitoring data, the method further includes: storing the aligned dietary data, medication data, exercise data and blood glucose monitoring data in a cloud database to establish a user data file in the cloud database.

[0011] According to the real-time intervention method provided by the present invention, the method further includes: Constructing a multi-dimensional and multi-level user health portrait based on the user's diet data, medication data, exercise data, and blood sugar monitoring data; The user portrait data is used to analyze user preference types, wherein the user preference types include: health management type, diet-oriented type or exercise-oriented type.

[0012] According to the real-time intervention method provided by the present invention, generating a real-time intervention suggestion according to the first blood glucose level prediction value and the second blood glucose level prediction value specifically includes: A behavioral intervention suggestion for the user is generated based on the first blood glucose level prediction value, the second blood glucose level prediction value, and the user preference data.

[0013] The present invention also provides a real-time intervention system based on blood glucose prediction of diabetic patients, comprising: A user data collection module, which is used to obtain blood glucose monitoring data and multi-dimensional life monitoring data based on blood glucose monitoring devices and smart devices, and to align the timestamps of the multi-dimensional life monitoring data with the blood glucose monitoring data; A blood glucose model prediction module, used for inputting the aligned multi-dimensional life monitoring data and blood glucose monitoring data into a pre-trained blood glucose prediction model, and outputting a first blood glucose level prediction value and a second blood glucose level prediction value, wherein the prediction period of the first blood glucose level prediction value is smaller than the prediction period of the second blood glucose level prediction value; the blood glucose prediction model is trained based on blood glucose monitoring sample data and life monitoring sample data of diabetic patients; An intervention suggestion proposing module, configured to generate a real-time intervention suggestion based on the first blood glucose level prediction value and the second blood glucose level prediction value; The hypoglycemia warning execution module is used to implement a graded warning if the first blood glucose level prediction value or the second blood glucose level prediction value is lower than a threshold value, and trigger the corresponding smart device to execute corresponding response measures.

[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the real-time intervention method based on blood glucose prediction of diabetic patients as described above is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described real-time intervention methods based on blood glucose prediction for diabetic patients.

[0016] The present invention also provides a computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the real-time intervention method based on blood glucose prediction of a diabetic patient as described above is implemented.

[0017] The real-time intervention method and system based on blood glucose prediction for diabetic patients provided by the present invention can output a first blood glucose level prediction value and a second blood glucose level prediction value by inputting multi-dimensional life monitoring data and blood glucose monitoring data into a pre-trained blood glucose prediction model, and generate real-time intervention suggestions based on these accurate prediction values. Patients can obtain guidance on diabetes without having to have profound professional medical knowledge, which helps them better understand changes in physiological indicators and adjust their daily behaviors accordingly; and when the blood glucose level prediction value is lower than the threshold, a graded warning is implemented, and the corresponding smart device is triggered to execute response measures. This warning mechanism significantly improves the quality of life of diabetic patients and optimizes blood glucose control through its real-time, normalized and highly personalized intervention strategy, combined with the intelligent operation of smart devices, greatly improving the convenience and intelligence of real-time and normalized intervention in blood glucose prediction for diabetic patients. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 It is a flow chart of the real-time intervention method based on blood glucose prediction of diabetic patients provided by the present invention.

[0020] Figure 2 This is one of the structural schematic diagrams of the real-time intervention system based on blood sugar prediction for diabetic patients provided by the present invention.

[0021] Figure 3 This is the second structural schematic diagram of the real-time intervention system based on blood sugar prediction for diabetic patients provided by the present invention.

[0022] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention.

[0023] Reference numerals: 100: Personal data monitoring and integration module; 200: Blood glucose model prediction module; 300: Patient portrait construction module; 400: Intervention suggestion module; 500: Model update and intervention suggestion module; 600: Hypoglycemia warning execution module; 700: user data collection module; 800: smart home appliance module. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are 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.

[0025] The method of the embodiment of the present invention obtains the first blood glucose level prediction value and the second blood glucose level prediction value through the blood glucose prediction model, and then generates real-time intervention suggestions, thereby having significant advantages in terms of accuracy of blood glucose prediction, real-time and personalization of intervention suggestions, etc., and can provide diabetic patients with a more scientific and effective blood glucose management plan, filling the deficiencies of the existing technology.

[0026] The embodiment of the present invention discloses a real-time intervention method based on blood glucose prediction of diabetic patients, which is used in a real-time intervention system. Figure 1 ,include: 101. Based on a blood glucose monitoring device and a smart device, obtain blood glucose monitoring data and multi-dimensional life monitoring data, and align timestamps of the multi-dimensional life monitoring data and the blood glucose monitoring data.

[0027] For blood glucose monitoring data: Device type: Use invasive or non-invasive continuous glucose monitoring (CGM) devices, such as Abbott Freestyle Libre, Dexcom G6, etc. These devices can monitor the patient's blood sugar level in real time and continuously, usually recording blood sugar values ​​every 5 minutes.

[0028] Data content: blood glucose value, timestamp, device status (such as whether the sensor is working properly).

[0029] Data transfer: Blood glucose data is transferred to an app on a smartphone or tablet via Bluetooth (BLE), Near Field Communication (NFC) or WiFi.

[0030] The multi-dimensional life monitoring data mainly includes the following: 1) Dietary data: Collection method: Patients input the time, type of food, and amount of each meal through a smartphone application. This can be done by text input, voice input, or taking a photo and uploading it. The image recognition function in the application can automatically identify the type of food and estimate the amount of intake.

[0031] Data content: meal time, food type, intake, nutritional components (such as calories, protein, fat, carbohydrates, etc.).

[0032] 2) Sports data: Device type: smart watch / bracelet, such as Apple Watch, Huawei smart watch, Garmin smart watch, etc.

[0033] Data content: number of steps, heart rate, calorie consumption, exercise duration, exercise intensity, exercise type (such as walking, running, swimming, etc.).

[0034] Data transfer: Sync your exercise data to the app on your smartphone or tablet via Bluetooth.

[0035] 3) Medication data: Collection method: Patients record medication information through smartphone applications, including drug name, medication time, dosage, and frequency. The information can be entered by scanning the drug QR code, and the patient can fill in the specific dosage and time of administration.

[0036] Data content: drug name, medication time, dosage, frequency, drug type (such as oral hypoglycemic drugs, insulin, etc.).

[0037] After obtaining blood glucose monitoring data and multi-dimensional life monitoring data, timestamp alignment is also required.

[0038] Data preprocessing: Check whether the blood glucose value is within a reasonable range (such as 20-600 mg / dL), remove obvious abnormal values ​​and mark possible sensor failure records. Screen extreme values ​​in exercise data (such as heart rate far exceeding the physiological limit or abnormal sudden increase in step count), and mark missing data. If necessary, simple missing value interpolation can be performed (such as linear interpolation using two valid measurements before and after) to ensure the consistency of the data time series.

[0039] Then feature extraction is performed to obtain data from multiple dimensions, including but not limited to time window features, derived mathematical features, diet features, exercise features, and medication impact features. For example, the average blood sugar value in the past hour, the total number of steps in the past 24 hours, the average medication dosage in the past 7 days, etc. are extracted.

[0040] During the timestamp alignment process, the timestamps of the multi-dimensional life monitoring data and the blood glucose monitoring data are aligned to ensure the time consistency of all data points. If the sampling rate of the CGM is uneven, linear interpolation is used to fill in the gaps in the data to ensure that the frequency of the data points input to the model is uniform.

[0041] The aligned data is stored in the cloud database to establish the user's data profile. The data profile includes blood glucose monitoring data, diet data, exercise data, and medication data, providing basic support for subsequent blood glucose prediction model training and real-time intervention recommendation generation.

[0042] In actual application, patients wear CGM devices and smart watches / bracelets in their daily lives and record diet and medication information through smartphone applications. The system automatically collects and transmits this data and aligns the timestamps.

[0043] When the patient is out, the system synchronizes the exercise data to the smartphone in real time through the Bluetooth function of the smart watch / bracelet. After returning home, all the data is uploaded to the cloud database via WiFi.

[0044] When a patient experiences symptoms of hypoglycemia, the system alerts the patient through the vibration function of the smart watch / bracelet and notifies the patient's family or emergency contacts through a smartphone app.

[0045] Through the extended implementation example content of the above step 101, the system can comprehensively and accurately collect and process the patient's blood glucose monitoring data and multi-dimensional life monitoring data, providing a solid foundation for subsequent blood glucose prediction and real-time intervention.

[0046] 102. Input the aligned multi-dimensional life monitoring data and blood glucose monitoring data into a pre-trained blood glucose prediction model, and output a first blood glucose level prediction value and a second blood glucose level prediction value.

[0047] Format the aligned multi-dimensional life monitoring data (diet, exercise, medication) and blood glucose monitoring data to ensure data consistency and readability. For example, convert all timestamps into a unified format and convert food types and intake in diet data into standardized nutritional parameters (such as calories, protein, fat, carbohydrates, etc.).

[0048] The formatted data are merged into a unified data set, with each data point containing timestamp, blood glucose value, diet information, exercise information, and medication information.

[0049] The blood glucose prediction model is trained based on the blood glucose monitoring sample data and life monitoring sample data of diabetic patients.

[0050] The model architecture uses LSTM based on attention mechanism: It uses the architecture of long short-term memory network (LSTM) combined with attention mechanism. LSTM can process time series data and capture the long-term dependencies of blood sugar changes, while the attention mechanism can dynamically weigh the importance of different input features and focus on the most relevant information for predicting future blood sugar levels.

[0051] Before entering the model, the data is feature engineered to extract useful features, such as the average blood sugar value in the past hour, the total number of steps in the past 24 hours, the average medication dosage in the past 7 days, etc.

[0052] During the model training process, a portion of all available patient data (e.g., 20 subjects) is selected to build a population model. The purpose of this stage is to allow the model to learn to capture the common blood sugar fluctuations in patients with type 2 diabetes. By training on a larger data set, the model can learn the common blood sugar response patterns among different patients.

[0053] Based on the population model, deep transfer learning technology is used to fine-tune the model parameters to adapt to the specific data of new patients (e.g., another 10 subjects). In this step, the model is trained using the initial data of the new patients (e.g., the data from the previous 1-2 days) in order to adjust the model to better adapt to the unique metabolic characteristics and lifestyle of individual patients.

[0054] By gradually increasing the scope of training data (for example, increasing it day by day), the model's adaptability to new patient data continues to be optimized. In this process, the model can gradually learn and incorporate more individual-specific information, thereby improving the individualized accuracy of the prediction.

[0055] After training, the model is evaluated on an independent test set, which usually includes subsequent time periods of patient data that were not used for training. This step is used to verify the model's predictive ability and generalization, ensuring that the model is not only applicable to patients in the training set, but can also effectively predict blood glucose levels for other patients.

[0056] The prediction output of the model includes short-term prediction and long-term prediction. The short-term prediction obtains the first blood sugar level prediction value, and the long-term prediction obtains the second blood sugar level prediction value.

[0057] Short-term prediction: The model can predict blood glucose levels within the next 30 minutes. For example, if the current time is 10:00 am, the model predicts that the blood glucose level at 10:30 will be 8.5 mmol / L.

[0058] Long-term prediction: The model can predict blood glucose levels within the next 60 minutes. For example, if the current time is 10:00 am, the model predicts that the blood glucose value at 11:00 will be 7.8 mmol / L.

[0059] The predictions are output as continuous values ​​that represent estimated blood glucose levels at specified future time points. The predictions are presented to patients via an interface that can be integrated into mobile applications to provide real-time, actionable blood glucose predictions.

[0060] In actual application, the system automatically collects data every 5 minutes and inputs the data into the pre-trained blood sugar prediction model to output short-term and long-term blood sugar level predictions in real time. For example, after a patient eats, the system will predict the blood sugar level in the next 30 minutes and 60 minutes based on their dietary data, recent blood sugar change trends, and blood sugar responses in similar situations in the past.

[0061] The model dynamically updates predictions based on the latest health data to reflect any potential blood sugar trends. For example, if a patient exercises immediately after eating, the model will adjust the prediction based on the new exercise data to provide a more accurate blood sugar prediction.

[0062] The system allows patients to provide feedback on the prediction results, such as "accurate prediction" or "inaccurate prediction". These feedback data will be used to further optimize the model and improve the accuracy and reliability of the prediction.

[0063] 103. Generate a real-time intervention suggestion based on the first blood glucose level prediction value and the second blood glucose level prediction value.

[0064] Intervention recommendations may include setting target blood glucose values. For example, short-term target blood glucose values: set short-term target blood glucose values ​​for the next 30 minutes, 60 minutes, 120 minutes, and 240 minutes based on the patient's treatment goals and actual conditions. The target blood glucose value allows a certain range of fluctuations to increase the possibility of obtaining a feasible solution. For example, for general patients, the target blood glucose value can be set at 6.0 - 7.0 mmol / L, allowing a fluctuation of ±0.5 mmol / L; for patients with poor blood glucose control or older patients, the target blood glucose value can be set at 7.0 - 8.0 mmol / L, allowing a fluctuation of ±1.0 mmol / L.

[0065] Long-term target blood glucose level: Set long-term target blood glucose levels for the next 7, 15, and 30 days for the design of overall intervention recommendations. For example, the long-term target blood glucose level can be set as HbA1c < 7.0%.

[0066] Intervention recommendations may also include dietary intervention, exercise intervention, and medication intervention. For example: For dietary intervention, dietary adjustment recommendations are generated based on the current blood sugar level, predicted blood sugar trend and target blood sugar value. For example, if the prediction shows that blood sugar will be high in 30 minutes, the patient is advised to reduce carbohydrate intake; if the prediction shows that blood sugar will be low, the patient is advised to increase the intake of sugary foods.

[0067] Specific recommendations: Generate specific dietary recommendations, such as "It is recommended that you reduce your carbohydrate intake by 50 grams at your next meal" or "It is recommended that you consume 15 grams of glucose within 30 minutes after a meal."

[0068] In addition, for exercise intervention, exercise adjustment suggestions are generated based on the current blood sugar level, predicted blood sugar change trend and target blood sugar value. For example, if the prediction shows that blood sugar will be high in 30 minutes, the patient is advised to do 30 minutes of moderate intensity exercise; if the prediction shows that blood sugar will be low, the patient is advised to reduce the intensity of exercise or suspend exercise.

[0069] Specific recommendations: Generate specific exercise recommendations, such as "It is recommended that you take a 30-minute brisk walk 30 minutes after a meal" or "It is recommended that you pause your exercise and rest for 15 minutes."

[0070] In addition, for medication intervention, medication adjustment suggestions are generated based on the current blood sugar level, predicted blood sugar change trend and target blood sugar value. For example, if the prediction shows that blood sugar will be high in 30 minutes, the patient is advised to increase the dose of oral hypoglycemic drugs; if the prediction shows that blood sugar will be low, the patient is advised to reduce the dose of insulin.

[0071] Specific recommendations: Generate specific medication recommendations, such as "It is recommended that you add 1 tablet of Metformin to your next medication dose" or "It is recommended that you reduce your next insulin injection by 2 units."

[0072] Opportunities for intervention may include the following: Regular intervention: Under routine monitoring, intervention suggestions are provided every 1-2 hours based on real-time data. For example, the system automatically generates intervention suggestions every hour and pushes them to patients through smartphone applications.

[0073] Abnormal fluctuation intervention: When abnormal fluctuations in the patient's blood sugar are monitored, additional intervention suggestions are immediately provided. For example, if the forecast shows that blood sugar will be lower than 3.9 mmol / L in 30 minutes, hypoglycemia warning and intervention suggestions are immediately generated and pushed.

[0074] Patient-demand intervention: When patients actively request intervention, additional intervention suggestions are immediately provided. For example, when a patient requests dietary suggestions through a smartphone app, the system immediately generates and pushes the corresponding dietary suggestions.

[0075] The push methods can include the following: Text prompts: intervention suggestions are pushed in text form through smartphone applications, such as "It is recommended that you reduce your carbohydrate intake by 50 grams in your next meal."

[0076] Voice reminders: Use smart watches / bracelets or smartphone apps to remind patients in voice form, such as "You are about to leave home, it is recommended that you bring an appropriate amount of sugar supplies to prevent hypoglycemia."

[0077] Push notifications: Push notifications are used to send intervention recommendations to patients and their families through push notification services (such as Firebase Cloud Messaging or Apple PushNotification) to ensure that they receive the recommendations in a timely manner and take appropriate actions.

[0078] 104. If the first blood glucose level predicted value or the second blood glucose level predicted value is lower than a threshold, implement a graded warning and trigger a corresponding smart device to execute a corresponding response measure.

[0079] Hypoglycemia warning trigger conditions: Short-term warning threshold, blood sugar can be set below 3.9 mmol / L after 30 minutes as the short-term hypoglycemia warning threshold; long-term warning threshold, blood sugar can be set below 3.9 mmol / L after 60 minutes as the long-term hypoglycemia warning threshold.

[0080] Short-term warning: If the first blood glucose level predicted value (after 30 minutes) is lower than 3.9 mmol / L, a short-term hypoglycemia warning is triggered.

[0081] Long-term warning: If the second blood glucose level predicted value (after 60 minutes) is lower than 3.9 mmol / L, a long-term hypoglycemia warning is triggered.

[0082] In addition, the graded warnings include: If the second predicted value of the blood sugar level is lower than the threshold or the first predicted value of the blood sugar level is lower than the threshold, a graded warning is made according to the distance between the user's current location and the residence: if the user's current location is at the residence, a first-level warning is made through the smart device, and the user's family is informed by message push; if the distance between the user's current location and the residence is less than or equal to the distance threshold, a second-level warning is made through the smart device, and the user's family is informed by message push, and the user's current location is shared in real time; if the distance between the user's current location and the residence is greater than the distance threshold, a third-level warning is made through the smart device, and the user's family is notified by phone, and the user's current location is shared in real time.

[0083] For example, for a Level 1 alert: Condition: The user's current location is in the place of residence.

[0084] Measures: Use smart watches / bracelets to vibrate slightly to remind patients; use smartphone apps to push hypoglycemia warning information to patients' families, informing them that the patient is at imminent risk of hypoglycemia.

[0085] For Level 2 warning: Condition: The distance between the user's current location and the place of residence is less than or equal to a 15-minute walking distance.

[0086] Measures: Use smart watches / bracelets to vibrate slightly to remind patients; use smartphone apps to push hypoglycemia warning information to patients’ families, and share the patient’s current location in real time, so that family members can quickly rush to the scene to provide assistance.

[0087] For Level 3 warning: Condition: The distance between the user's current location and the place of residence is greater than a 15-minute walking distance.

[0088] Measures: Use smart watches / bracelets to vibrate strongly to remind patients and emphasize the urgency of the situation; use smart watches / bracelets or smartphone apps to make calls directly to the patient's family and share the patient's real-time location; use smart devices to play pre-recorded help voices to remind people around to provide help.

[0089] The real-time intervention method based on blood glucose prediction for diabetic patients provided by the embodiment of the present invention can output a first blood glucose level prediction value and a second blood glucose level prediction value by inputting multi-dimensional life monitoring data and blood glucose monitoring data into a pre-trained blood glucose prediction model, and generate real-time intervention suggestions based on these accurate prediction values. Patients do not need to have profound professional medical knowledge to obtain guidance on diabetes, which helps them better understand the changes in physiological indicators and adjust their daily behaviors accordingly; and when the blood glucose level prediction value is lower than the threshold, a graded warning is implemented, and the corresponding smart device is triggered to execute response measures. This warning mechanism significantly improves the quality of life of diabetic patients and optimizes blood glucose control through its real-time, normalized and highly personalized intervention strategy, combined with the intelligent operation of smart devices, greatly improving the convenience and intelligence of real-time, normalized intervention in blood glucose prediction for diabetic patients.

[0090] Furthermore, if the user is detected to be in a hypoglycemic state based on the blood sugar monitoring data and the multi-dimensional life monitoring data, at least one of the following processing measures is performed: Contact the user's family via phone and share the user's real-time location; Play voice messages through smart devices to alert people around to provide help; If it detects that there is no one around the user, it will automatically call the emergency number and explain the user's situation and location through the intelligent voice system embedded in the smart device.

[0091] Emergency measures include the following: Contacted family members by phone: The patient's blood sugar was detected to be lower than 3.9 mmol / L, and the patient was in a state of hypoglycemia.

[0092] Measures: Automatically dial the phone: Automatically dial the phone to the patient’s emergency contacts (such as family members and friends) through the communication module of the smartphone application or smart watch / bracelet.

[0093] Share location information: After the call is placed, the patient’s real-time location information is sent to emergency contacts via SMS or in-app messaging to ensure they can find the patient quickly.

[0094] Voice broadcast: After the call is connected, the text-to-speech system (TTS) in the smart device broadcasts the patient's current hypoglycemia status and location information, for example: "Patient Zhang Ming's current blood sugar level is 3.5 mmol / L, and his location is No. XX, XX Street. He needs emergency assistance." A voice message is played to remind people around: the patient's blood sugar is detected to be lower than 3.9 mmol / L and the patient is in a state of hypoglycemia.

[0095] Measures: Voice reminder: Play a pre-recorded help voice through the speaker of a smart watch / bracelet or smartphone to remind people around the patient of the current hypoglycemia state, for example: "I am a diabetic patient, my blood sugar is currently too low, and I need emergency assistance. Please help me." Repeat play: The help voice message will be played repeatedly every 30 seconds until the patient or people around take action.

[0096] Automatically dial emergency number: detect that the patient's blood sugar is lower than 3.9 mmol / L and the patient is in a state of hypoglycemia, and detect that there is no one around the patient (determined by the sound wave detection or sensor of the smart device).

[0097] Measures: Automatically dial 120: Automatically dial the emergency number 120 through the communication module of the smartphone application or smart watch / bracelet.

[0098] Voice instructions: After the call is connected, the voice synthesis system (TTS) in the smart device will explain the patient's current hypoglycemia status and location information to the emergency operator, for example: "This is the smart health management system. The patient Zhang Ming's current blood sugar level is 3.5 mmol / L. He is located at No. XX, XX Street, and needs emergency medical assistance." Continuous communication: Maintain communication with the emergency center until emergency personnel arrive on the scene.

[0099] For example, for actual application scenarios: Home environment: The patient suddenly felt dizzy and palpitated at home, and the blood sugar monitoring device showed that the blood sugar was 3.2 mmol / L.

[0100] Smart watch / bracelet vibration reminder: The smart watch / bracelet vibrates slightly to remind the patient of hypoglycemia. Contact family members by phone: The system automatically calls the patient's family members and shares the patient's real-time location information via text message. Voice reminder: The smart watch / bracelet plays a pre-recorded voice for help to remind family members or cohabitants to provide help.

[0101] Outdoor environment: The patient suddenly felt dizzy and panicky when going out. The blood sugar monitoring device showed that the blood sugar was 3.2mmol / L, and there was no one around.

[0102] Smart watch / bracelet vibration reminder: The smart watch / bracelet vibrates strongly to remind the patient of hypoglycemia; Contact family members by phone: The system automatically calls the patient's family and shares the patient's real-time location information via text message; Automatically dial emergency numbers: When the system detects that there is no one around the patient, it automatically dials 120 and explains the patient's condition and location to the emergency operator through the speech synthesis system; Voice reminder: The pre-recorded help voice is played through the smart watch / bracelet to remind people around to provide help.

[0103] Public places: The patient suddenly felt dizzy and had palpitations in public places such as shopping malls and parks, and the blood sugar monitoring device showed that the blood sugar was 3.2 mmol / L.

[0104] Smart watch / bracelet vibration reminder: The smart watch / bracelet vibrates slightly to remind the patient of hypoglycemia; play voice message: play pre-recorded help voice through the smart watch / bracelet to remind people around to provide help; contact family by phone: the system automatically calls the patient's family and shares the patient's real-time location information via text message; automatically dial emergency number: if people around cannot provide help in time, the system automatically dials 120 and explains the patient's condition and location to the emergency operator through the voice synthesis system.

[0105] Through the contents of the above embodiments, the system of this embodiment can take a variety of emergency treatment measures in a timely manner when it detects that the patient is in a state of hypoglycemia, ensure the safety of the patient in hypoglycemia risks and events, and provide timely rescue and support.

[0106] Furthermore, after triggering the corresponding smart device to execute corresponding response measures, the method also includes: triggering smart home appliances connected to the smart device to execute corresponding response measures; wherein the smart home appliances include at least one of the following: smart door locks, smart refrigerators, and smart cookers.

[0107] Specifically, the smart door lock can remind the patient before leaving home: when the patient is about to leave home, based on the blood sugar prediction results, the voice reminder function of the smart door lock can be used to remind the patient to bring sugar supplies, and the smart refrigerator can be linked to provide specific supply suggestions through the display screen.

[0108] Smart door locks can be used for emergency rescue: when a patient is detected to have severe hypoglycemia and no one is at home, the high-resolution camera built into the smart door lock automatically starts to scan the identity of the visitor in real time. If the person is identified as an emergency personnel, the smart door lock automatically unlocks and allows him to enter the house for rescue.

[0109] Smart refrigerators can make food recommendations: based on the patient's dietary intervention recommendations and the food inventory in the refrigerator, personalized recommended recipes and cooking methods are generated and displayed simultaneously through the smart refrigerator's touch screen and the patient's smartphone application.

[0110] Smart refrigerators can achieve inventory management: through built-in high-definition cameras and weight sensors, they automatically record the storage and access of food, ensure the accuracy of the records, and upload the data to the cloud database.

[0111] Automatic cooking by smart cookers: When patients choose to accept the recipes recommended by the refrigerator, the smart cookers can automatically set the cooking parameters and execute the preset cooking process, reducing the patient's operational burden and ensuring the effective implementation of dietary recommendations.

[0112] The system of this embodiment can trigger graded warnings in a timely manner when hypoglycemia risks are predicted, and execute corresponding response measures through a variety of smart devices and smart home appliances to ensure the safety of patients in hypoglycemia risks and events, and provide timely assistance and support.

[0113] In addition, after aligning the timestamps of the multi-dimensional life monitoring data and the blood glucose monitoring data, the method further includes: storing the aligned diet data, medication data, exercise data and blood glucose monitoring data in a cloud database to establish a user data file in the cloud database.

[0114] Cloud databases can use relational databases (such as MySQL, PostgreSQL) or non-relational databases (such as MongoDB, Cassandra) to store user data files. The database design should include the following table structure: User table: stores basic user information, such as user ID, name, age, gender, medical history, etc.

[0115] Blood glucose monitoring data table: stores blood glucose monitoring data, including timestamp, blood glucose value, device ID, etc.

[0116] Diet data table: stores diet data, including timestamp, food type, intake, nutritional content, etc.

[0117] Exercise data table: stores exercise data, including timestamp, number of steps, heart rate, calorie consumption, exercise duration, exercise intensity, exercise type, etc.

[0118] Medication data table: stores medication data, including timestamp, drug name, dosage, frequency, drug type, etc.

[0119] The aligned multi-dimensional life monitoring data and blood glucose monitoring data are stored in the cloud database to ensure data integrity and consistency.

[0120] Furthermore, the method also includes: constructing a multi-dimensional, multi-level user health portrait based on the user's diet data, medication data, exercise data and blood sugar monitoring data; and analyzing user preference types based on the user portrait data, wherein the user preference types include: health management type, diet-dominated type or exercise-dominated type.

[0121] A behavioral intervention suggestion for the user is generated based on the first blood glucose level prediction value, the second blood glucose level prediction value, and the user preference data.

[0122] Analysis of user preference types, including: Health management type: Patients have good self-management abilities in diet, exercise, medication, etc., and pay attention to maintaining overall health status.

[0123] Diet-dominated type: Patients have strong preferences in diet, their eating habits have a greater impact on blood sugar control, and their compliance with dietary adjustments is high.

[0124] Exercise-dominant type: Patients have a strong preference for exercise, exercise habits have a greater impact on blood sugar control, and compliance with exercise adjustments is high.

[0125] By analyzing the user's health profile data, the user's preference type can be identified. For example, if the patient's diet data occupies an important position in the health profile and dietary adjustment has a significant effect on blood sugar control, it is identified as diet-dominated.

[0126] Combined with user feedback, the user's preference type is further confirmed and refined. For example, if a user expresses in the app that he or she would rather adjust his or her diet than increase exercise, the system will identify the user as a diet-oriented person.

[0127] The real-time intervention system provided by the present invention is described below. The real-time intervention system described below and the real-time intervention method described above can be referenced to each other.

[0128] See also Figure 2 , Figure 2 A schematic structural diagram of the real-time intervention system of this embodiment is shown.

[0129] The user data collection module 700 is used to obtain blood glucose monitoring data and multi-dimensional life monitoring data based on the blood glucose monitoring device and the smart device, and align the timestamps of the multi-dimensional life monitoring data with the blood glucose monitoring data; The blood glucose model prediction module 200 is used to input the aligned multi-dimensional life monitoring data and the blood glucose monitoring data into a pre-trained blood glucose prediction model, and output a first blood glucose level prediction value and a second blood glucose level prediction value, wherein the prediction period of the first blood glucose level prediction value is smaller than the prediction period of the second blood glucose level prediction value; the blood glucose prediction model is trained based on the blood glucose monitoring sample data and the life monitoring sample data of the diabetic patient; An intervention suggestion proposing module 400 is used to generate a real-time intervention suggestion according to the first blood glucose level prediction value and the second blood glucose level prediction value; For an explanation of the specific real-time intervention suggestions, please refer to the contents of the aforementioned embodiments, which will not be repeated here.

[0130] The hypoglycemia warning execution module 600 is used to implement a graded warning and trigger a corresponding smart device to execute a corresponding response measure if the first blood glucose level prediction value or the second blood glucose level prediction value is lower than a threshold value.

[0131] In addition to the above functions, the functions of the remaining modules are as follows: User data collection module (700): This module collects the patient's blood sugar, exercise, diet and medication data in real time through the front-end device, and transmits it to the cloud securely and efficiently.

[0132] Data collection from invasive or non-invasive CGM devices: Wear / attach an invasive CGM sensor (e.g., subcutaneous sensor) or non-invasive CGM device to the patient and position it correctly according to the device manual. After starting the CGM device, pair the device with the smartphone app via Bluetooth (BLE) or near-field communication (NFC). The CGM device automatically measures blood glucose at predetermined intervals (e.g., every 5 minutes) and sends the readings to the smartphone. The mobile app structures the data locally (e.g., in JSON format) and performs preliminary validation (checking the value range, timestamp integrity).

[0133] Smart watch / bracelet motion data collection: Wear the smart watch / bracelet on your wrist as instructed so that the sensors (accelerometer, heart rate sensor, etc.) can stably capture human activity signals. Use a smartphone application to pair the watch / bracelet with Bluetooth. The application will periodically (e.g., every 10 minutes) request data synchronization from the watch / bracelet in the background. The watch / bracelet will return a set of motion parameters (number of steps, heart rate, calorie consumption, exercise duration, etc.) and a timestamp. The application parses and verifies the data (e.g., screens out outliers) and uploads it to the cloud via HTTPS. When location information is needed (e.g., for hypoglycemia rescue positioning), the application can call the mobile phone GPS function and periodically upload the location information to the cloud on the premise of ensuring user authorization to provide a geographic reference for subsequent decision-making.

[0134] Diet and medication data entry on smartphones or mobile terminals: Users enter diet and medication information through a dedicated interface in smartphone applications. For dietary data, text input (describing meal contents), voice input (voice recognition text), or photo upload (converted into nutritional parameters through image recognition) can be used. For medication information, users can enter the name of the drug, dosage, and medication time, or take a photo of the medicine box barcode and the application will automatically complete the information by retrieving the drug database. After the input is completed, the application performs data compliance verification locally (such as checking whether the dosage is a positive number and whether the date and time format is correct), and the data is uploaded to the cloud server via HTTPS encryption. If the network is interrupted, the data is stored in a local encrypted file and uploaded by the background synchronization thread after the network is restored.

[0135] Personal data monitoring and integration module (100): This module performs standardization, data cleaning and correlation on the data uploaded by module 700 in the cloud, providing high-quality input for the prediction and intervention module. The specific implementation is as follows: Continuous blood glucose monitoring data integration: The backend data processing program analyzes and verifies the blood glucose value and the corresponding timestamp: Checks whether the blood glucose value is within a reasonable range (such as 20-600 mg / dL), removes obvious abnormal values ​​and marks possible sensor failure records. The cleaned blood glucose data is uniformly stored in the blood glucose time series data table of the backend database, and indexed by user ID and timestamp, providing continuous and complete time series input for subsequent blood glucose prediction model construction and analysis.

[0136] Sports monitoring data integration: Receive sports data uploaded from smart watches / bracelets, including steps, heart rate, calorie consumption, and exercise duration. Format the sports data, map each parameter to predefined data fields (such as "heart_rate", "steps", "calories", "duration"), and match the corresponding timestamp and user ID. Filter extreme values ​​(such as heart rate far exceeding the physiological limit or abnormal sudden increase in step count), and mark missing data. If necessary, simple missing value interpolation can be performed (such as linear interpolation using two valid measurements before and after) to ensure the consistency of the data time series. The cleaned sports data is stored in the sports index data table and associated with the blood sugar data through the user ID and the similar time window, providing data support for analyzing the relationship between exercise behavior and blood sugar fluctuations.

[0137] Diet monitoring data integration: Receive diet diary text data and food image recognition results from a smartphone or mobile terminal. The text data has been verified locally by the application when it is uploaded, and the food images are processed by cloud-based image recognition and nutrition analysis services to return the calorie value and nutritional composition of each dish. Match the food entries and portion information in the diary text records with the nutritional data returned by image recognition, and construct a structured diet data record through standardized rules (such as unified calorie units, marked protein, fat, and carbohydrate mass values). Mark or remove outliers (such as extremely high calorie intake that does not conform to contextual logic), store the cleaned diet data in the diet data table, and use timestamps and user IDs as key values ​​to achieve retrievable multi-dimensional associations with blood sugar and exercise data.

[0138] Medication monitoring data integration: Receive medication information entered by users on the mobile terminal, including drug name, dosage, medication time and frequency. Compare the drug name with the local or cloud drug information database to determine the drug ID and standard dosage unit. Mark and review abnormal records (such as drug names that do not exist in the drug database or dosage values ​​that are significantly beyond the reasonable range). Incorporate medication data into the integration framework through timestamps. The cleaned medication records are stored in the medication data table, and through the time matching mechanism with blood glucose data, a traceable data basis is provided for subsequent analysis of the impact of medication regimens on blood glucose control.

[0139] Blood glucose model prediction module (200): This module is divided into two submodules: building a personal blood glucose metabolism model and predicting blood glucose at future moments.

[0140] Model construction: Collect and integrate patients' health data and perform feature engineering to improve the accuracy of model training. Machine learning algorithms (such as transformer, LSTM, SVM, etc.) can be used for model training, or mathematical, physiological and other modeling methods can be used to build an optimized blood sugar metabolism model.

[0141] Blood sugar prediction: Input the latest health monitoring data in real time and dynamically update the prediction results. Output short-term (such as 30 / 60 / 120 / 240 minutes) and long-term (such as 7 / 15 / 30 days) blood sugar predictions to support real-time and long-term diabetes management.

[0142] Patient portrait construction module (300): constructs a multi-dimensional health portrait through comprehensive analysis of the patient's basic information and monitoring data.

[0143] Blood sugar dimension: Classify patients’ blood sugar control levels and evaluate the effectiveness of blood sugar management.

[0144] Diet, exercise and medication dimensions: Analyze the patient's eating habits, daily activity levels and medication patterns to form a comprehensive health portrait.

[0145] Intervention suggestion proposing module (400): This module combines blood glucose prediction with user preferences to dynamically generate intervention suggestions.

[0146] Goal setting and recommendation generation: Set real-time target blood sugar values ​​based on the patient's condition, use algorithms to traverse and analyze the possibilities of different diet and exercise combinations, and combine patient profiles to screen feasible types of recommendations and recommended doses.

[0147] Human Selection and Optimization: Professionals review and adjust recommendations to ensure practicality and feasibility.

[0148] Model update and intervention suggestion module (500): This module optimizes the formulation strategy of personal blood glucose model and intervention suggestion through patient feedback and blood glucose monitoring data.

[0149] Compliance and satisfaction tracking: Monitor patients’ implementation and satisfaction with intervention recommendations to adjust subsequent recommendation strategies.

[0150] Continuous iteration and feedback: The system continuously learns and optimizes subsequent intervention measures to improve patient satisfaction and blood sugar control.

[0151] Specifically, the connection relationship between modules 200, 300, 400 and 500 is as follows: Figure 3 shown.

[0152] Initialization and model application: The process of proposing each intervention recommendation begins with the initialization of the system, at which time the system has loaded the patient's basic information and historical medical data. The first step is to apply a personal blood glucose metabolism model, which combines real-time monitoring data with historical blood glucose data to predict the patient's blood glucose level at a specific time in the future (such as 30 minutes later).

[0153] Evaluation and intervention decision: The system next evaluates whether the predicted blood sugar value is within the predefined safety range. If the predicted value is within a reasonable range, the process will terminate and continue with regular blood sugar monitoring. If the predicted value exceeds the safety range, the system will look for the blood sugar value closest to the previous predicted value within the safety range, and reversely calculate the permutations and combinations of diet and exercise intervention doses that can make blood sugar reach the target value after a specific time in the future under the current blood sugar state. Next, based on the patient's personal health portrait, the type and dose of intervention recommendations suitable for the patient are eliminated and handed over to medical staff for appropriate and final decision.

[0154] Implementation and monitoring of intervention recommendations: Once intervention recommendations are generated, the system will convey these recommendations to patients through appropriate communication channels (such as text prompts or voice reminders) to guide them to adjust their daily diet and exercise behaviors. After that, the system will continue to use blood glucose monitoring equipment to track blood glucose changes after the intervention and collect relevant data.

[0155] Effect evaluation and optimization suggestions: After the intervention is implemented, the system will analyze the subsequent blood sugar data collected to evaluate the actual effect of the intervention. Based on the evaluation results of the intervention effect, the system may confirm the current intervention strategy or make necessary adjustments. Adjustments may include changing the type of diet, adjusting the intensity or duration of the exercise plan, etc., to ensure that patients can achieve better blood sugar control in subsequent management.

[0156] End and continuous monitoring: After the process is completed, the system will continue to monitor the patient's blood sugar status and be ready to restart the generation and optimization process of intervention suggestions based on new data input. This combination of phased and continuous ensures the continuity and real-time nature of the patient's blood sugar management and enhances the adaptability and effectiveness of the treatment plan.

[0157] Hypoglycemia warning execution module (600): This module implements graded warning and diversified rescue measures based on the constructed blood sugar prediction model (the blood sugar model prediction module 200 is responsible for outputting the prediction results) and real-time location information.

[0158] Method for determining the time threshold for hypoglycemia risk: If the predicted value is lower than the set threshold (such as 70 mg / dL), the corresponding warning process will be triggered based on the predicted time margin (30 minutes or 60 minutes).

[0159] 60-minute advance warning method: If hypoglycemia is predicted to occur in 60 minutes, the backend sends a warning instruction to the user's mobile phone through the message push service (Firebase Cloud Messaging or Apple Push Notification). After receiving the instruction, the mobile phone sends a vibration command to the bracelet via Bluetooth, activating the bracelet to vibrate slightly to remind the user to pay attention in advance.

[0160] 30-minute advance measures determination method: When the prediction result triggers a 30-minute advance warning, the system calculates the walking distance between the two points based on the user's location (obtained from the mobile phone's GPS data) and the coordinates of the place of residence. (1) If the user is at home or other fixed residence (distance is 0): the bracelet vibrates to remind the user, and the mobile phone message push API is used to notify the family application. After the family APP receives the message, they can view the prompt information; (2) If the distance is ≤15 minutes walking range: in addition to the above steps, the user's real-time location is also transmitted to the family APP, and the user's location is displayed on the map interface for quick response; (3) If the distance is >15 minutes walking range: in addition to vibration and message push, a call is made directly through the cloud (calling VoIP or PSTN gateway) to contact the family. After the call is connected, the LLM (large language model) and TTS (text to speech) broadcast the patient's current condition (including blood sugar, heart rate, etc.) and location coordinates to improve communication timeliness.

[0161] Emergency treatment when hypoglycemia has occurred: Real-time analysis of CGM data. If the current blood sugar value has fallen below the threshold, it will detect whether the user has fallen. The fall judgment is completed through the mobile phone acceleration sensor data and posture recognition algorithm: if the user lies flat for a long time after a large acceleration change, it is judged as a fall. If the person has not fallen: the family will be automatically contacted by phone and the location will be notified through LLM and TTS. The mobile phone or bracelet will play a pre-recorded help voice to prompt nearby people to help. If the person has fallen or is bedridden: the family will also be contacted and the location will be shared, and the sound wave detection technology (the microphone recording can be called in the mobile phone APP to judge the characteristics of the ambient sound or the reaction to the volume of other people's speech) will be used to determine whether there are people around. If there is someone, the help voice will be played again; if no one responds, the emergency number (120) will be automatically dialed, and the patient's condition and location will be described to the operator using LLM and TTS during the call.

[0162] Smart home appliance module (800): This module cooperates with modules 200, 400 and 600 to provide blood sugar management support for patients in a home environment. Specifically, it provides support in the following three smart appliance modes, but not limited to.

[0163] Smart door locks: Smart door locks integrate voice reminders, camera monitoring and automatic unlocking functions to achieve intelligent management of patients' behavior away from home and emergency rescue support.

[0164] Sugar replenishment reminder before leaving home: The system obtains the patient's blood sugar prediction data for the next 30 minutes and 60 minutes through the blood sugar model prediction module (200) to assess the patient's blood sugar risk after leaving home. If the prediction result shows that the patient may have a risk of hypoglycemia after leaving home, the system automatically generates a sugar replenishment reminder. The smart door lock plays a voice reminder through the built-in speaker: "You are about to leave home. It is recommended to bring an appropriate amount of sugar replenishment to prevent hypoglycemia." At the same time, the smart refrigerator combines the inventory of the contents and prompts specific replenishment suggestions through its display screen, such as: "It is recommended to bring a banana or a glass of low GI fruit juice." Rescue assistance in severe hypoglycemia events: When the linked hypoglycemia warning execution module (600) detects that a patient has a severe hypoglycemia event (such as falling or lying in bed) and no one is at home, the emergency rescue mode is automatically triggered. The high-resolution camera built into the smart door lock is automatically activated to scan the identity of the visitor outside the door in real time. The system uses image recognition technology and built-in voice interaction to determine the identity of the visitor and whether he or she is an emergency personnel. If identified as an emergency personnel, the smart door lock automatically unlocks and allows him or her to enter the home for rescue. If identified as a non-emergency personnel, the system keeps the door lock locked, and further notifies the medical team and family through module 600 to start the backup rescue process.

[0165] Smart refrigerator: The smart refrigerator automatically generates personalized recommended recipes and cooking methods by linking with module 400 based on dietary intervention recommendations and the type and quantity of food in the refrigerator, taking into account the patient's taste characteristics.

[0166] Food classification and knowledge graph construction: Using knowledge graph technology, the transaction information frequently purchased by patient users is divided into five levels according to the glycemic index. The classified food content data is constructed into a user-related knowledge graph and stored in a cloud database for subsequent query and analysis.

[0167] Automatic food recording and inventory management: The smart refrigerator has a built-in high-definition camera that uses image recognition technology to identify the type and quantity of food stored and accessed in real time. Optionally, a precise weight sensor is installed in the food storage area to measure the specific amount of food taken out or stored to ensure the accuracy of the record. When the patient takes out or stores food from the refrigerator, the camera and weight sensor work together to automatically identify the type, GI level and quantity of food. The system uploads the identified food data (food name, GI level, quantity, timestamp) to the cloud server through an encrypted protocol and stores it in association with the dietary monitoring data in the comprehensive personal data monitoring and integration module (100).

[0168] Dietary advice reading and recipe calculation and generation: Based on the dietary advice output by module 400, the smart refrigerator module queries the current food inventory information in the refrigerator, including the GI level and quantity of various foods. Combined with the actual inventory of food in the refrigerator, feasible food combinations are screened out, and combined with the user taste characteristics in the patient portrait, the operability and practicality of the recommended recipes are ensured. The recommended recipes and cooking methods are displayed to the patient through the touch screen of the smart refrigerator. Allow patients to choose to accept suggestions or make adjustments, such as changing ingredients, adjusting portions, etc., and call module 200 to recalculate and provide blood sugar prediction feedback for the adjusted recipes. Record the patient's selection and adjustment information to optimize the knowledge graph and recipe generation algorithm, and improve the accuracy and personalization of the suggestions.

[0169] Smart cookers: Smart cookers can achieve a more automated cooking process by linking with smart refrigerators.

[0170] Recipe reception: When the patient selects and confirms to accept the suggestion through the interactive interface of the smart refrigerator module, the system transmits the recipe information (including required ingredients, cooking methods, temperature, time, etc.) to the smart cooker module. The smart cooker analyzes the received recipe information and extracts key parameters such as cooking methods, temperature and time.

[0171] Automatic cooking settings: Based on the analysis results, smart cookers (such as smart ovens, smart rice cookers, etc.) automatically set the corresponding cooking parameters. For example, if the recipe recommends "roast chicken breast", the smart oven automatically adjusts to 180°C and sets the baking time to 30 minutes.

[0172] Cooking feedback: Patients can evaluate and provide feedback on cooking results through the application, such as cooking effect, taste satisfaction, etc. The system optimizes cooking parameters and recipe recommendation algorithms based on the feedback to improve the accuracy and personalization of future recipe suggestions.

[0173] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4As shown, the electronic device may include: a processor (processor) 410 , a communication interface (Communications Interface) 420 , a memory (memory) 430 and a communication bus 440 , wherein the processor 410 , the communication interface 420 , and the memory 430 communicate with each other through the communication bus 440 . The processor 410 can call the logic instructions in the memory 430 to execute a real-time intervention method based on blood glucose prediction for diabetic patients, the method comprising: based on blood glucose monitoring devices and smart devices, obtaining blood glucose monitoring data and multidimensional life monitoring data, and aligning the multidimensional life monitoring data with the blood glucose monitoring data by timestamp; inputting the aligned multidimensional life monitoring data and blood glucose monitoring data into a pre-trained blood glucose prediction model, and outputting a first blood glucose level prediction value and a second blood glucose level prediction value, wherein the prediction period of the first blood glucose level prediction value is less than the prediction period of the second blood glucose level prediction value; the blood glucose prediction model is trained based on the blood glucose monitoring sample data and life monitoring sample data of diabetic patients; generating real-time intervention suggestions based on the first blood glucose level prediction value and the second blood glucose level prediction value; if the first blood glucose level prediction value or the second blood glucose level prediction value is lower than the threshold, implementing a graded warning, and triggering the corresponding smart device to execute corresponding response measures.

[0174] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0175] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the real-time intervention method based on blood glucose prediction of diabetic patients provided by the above-mentioned methods, the method including: based on blood glucose monitoring equipment and smart devices, obtaining blood glucose monitoring data and multidimensional life monitoring data, and aligning the multidimensional life monitoring data with the blood glucose monitoring data by timestamp; inputting the aligned multidimensional life monitoring data and blood glucose monitoring data into a pre-trained blood glucose prediction model, and outputting a first blood glucose level prediction value and a second blood glucose level prediction value, wherein the prediction period of the first blood glucose level prediction value is less than the prediction period of the second blood glucose level prediction value; the blood glucose prediction model is trained based on the blood glucose monitoring sample data and life monitoring sample data of diabetic patients; generating real-time intervention suggestions based on the first blood glucose level prediction value and the second blood glucose level prediction value; if the first blood glucose level prediction value or the second blood glucose level prediction value is lower than the threshold, implementing a graded warning, and triggering the corresponding smart device to execute corresponding response measures.

[0176] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the real-time intervention method based on blood glucose prediction for diabetic patients provided by the above-mentioned methods, the method comprising: based on a blood glucose monitoring device and an intelligent device, obtaining blood glucose monitoring data and multidimensional life monitoring data, and aligning the multidimensional life monitoring data with the blood glucose monitoring data by timestamp; inputting the aligned multidimensional life monitoring data and blood glucose monitoring data into a pre-trained blood glucose prediction model, and outputting a first blood glucose level prediction value and a second blood glucose level prediction value, wherein the prediction period of the first blood glucose level prediction value is less than the prediction period of the second blood glucose level prediction value; the blood glucose prediction model is trained based on the blood glucose monitoring sample data and life monitoring sample data of the diabetic patient; generating real-time intervention suggestions based on the first blood glucose level prediction value and the second blood glucose level prediction value; if the first blood glucose level prediction value or the second blood glucose level prediction value is lower than a threshold, implementing a graded warning, and triggering the corresponding intelligent device to execute corresponding response measures.

[0177] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0178] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0179] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A real-time intervention method based on blood glucose prediction for diabetic patients, characterized in that: include: Based on the blood glucose monitoring device and the smart device, blood glucose monitoring data and multi-dimensional life monitoring data are obtained, and the multi-dimensional life monitoring data and the blood glucose monitoring data are time-stamped and aligned; Inputting the aligned multi-dimensional life monitoring data and blood glucose monitoring data into a pre-trained blood glucose prediction model, outputting a first blood glucose level prediction value and a second blood glucose level prediction value, wherein the prediction period of the first blood glucose level prediction value is shorter than the prediction period of the second blood glucose level prediction value; the blood glucose prediction model is trained based on blood glucose monitoring sample data and life monitoring sample data of diabetic patients; generating a real-time intervention recommendation based on the first blood glucose level prediction value and the second blood glucose level prediction value; If the first blood sugar level predicted value or the second blood sugar level predicted value is lower than a threshold, a graded warning is implemented, and the corresponding smart device is triggered to execute corresponding response measures.

2. The real-time intervention method according to claim 1, characterized in that: The method further includes: when it is detected according to the blood sugar monitoring data and the multi-dimensional life monitoring data that the user is in a hypoglycemic state, executing at least one of the following processing measures: Contact the user's family via phone and share the user's real-time location; Play voice messages through smart devices to alert people around to provide help; If it detects that there is no one around the user, it will automatically call the emergency number and explain the user's situation and location through the intelligent voice system embedded in the smart device.

3. The real-time intervention method according to claim 1, characterized in that: If the first blood glucose level prediction value or the second blood glucose level prediction value is lower than the threshold, a graded warning is implemented, and the corresponding smart device is triggered to execute corresponding response measures, specifically including: If the second predicted value of the blood sugar level is lower than the threshold or the first predicted value of the blood sugar level is lower than the threshold, a graded warning is issued according to the distance between the user's current location and the residence: If the user's current location is where they live, a first-level warning will be issued through the smart device, and the user's family will be notified through push notifications; If the distance between the user's current location and the place of residence is less than or equal to the distance threshold, a second-level warning will be issued through the smart device, and the user's family will be informed through message push, and the user's current location will be shared in real time; If the distance between the user's current location and residence is greater than the distance threshold, a third-level warning will be issued through the smart device, and the user's family will be notified by phone, and the user's current location will be shared in real time.

4. The real-time intervention method according to claim 1 or 3, characterized in that: After triggering the corresponding smart device to execute the corresponding response measures, the method further includes: Triggering the smart home appliances connected to the smart device to execute corresponding response measures; wherein the smart home appliances include at least one of the following: smart door locks, smart refrigerators, and smart cookers.

5. The real-time intervention method according to claim 1, characterized in that: Multi-dimensional life monitoring data includes: diet data, medication data, and exercise data; After aligning the timestamps of the multi-dimensional life monitoring data and the blood glucose monitoring data, the method further includes: The aligned diet data, medication data, exercise data and blood sugar monitoring data are stored in a cloud database to establish a user data file in the cloud database.

6. The real-time intervention method according to claim 5, characterized in that: The method further comprises: Constructing a multi-dimensional and multi-level user health portrait based on the user's diet data, medication data, exercise data, and blood sugar monitoring data; The user portrait data is used to analyze user preference types, wherein the user preference types include: health management type, diet-oriented type or exercise-oriented type.

7. The real-time intervention method according to claim 6, characterized in that: Generating a real-time intervention suggestion based on the first blood glucose level prediction value and the second blood glucose level prediction value, specifically comprising: A behavioral intervention suggestion for the user is generated based on the first blood glucose level prediction value, the second blood glucose level prediction value, and the user preference data.

8. A real-time intervention system based on blood sugar prediction for diabetic patients, characterized in that: include: A user data collection module, which is used to obtain blood glucose monitoring data and multi-dimensional life monitoring data based on blood glucose monitoring devices and smart devices, and to align the timestamps of the multi-dimensional life monitoring data with the blood glucose monitoring data; A blood glucose model prediction module, used for inputting the aligned multi-dimensional life monitoring data and blood glucose monitoring data into a pre-trained blood glucose prediction model, and outputting a first blood glucose level prediction value and a second blood glucose level prediction value, wherein the prediction period of the first blood glucose level prediction value is smaller than the prediction period of the second blood glucose level prediction value; the blood glucose prediction model is trained based on blood glucose monitoring sample data and life monitoring sample data of diabetic patients; An intervention suggestion proposing module, configured to generate a real-time intervention suggestion based on the first blood glucose level prediction value and the second blood glucose level prediction value; The hypoglycemia warning execution module is used to implement a graded warning if the first blood glucose level prediction value or the second blood glucose level prediction value is lower than a threshold value, and trigger the corresponding smart device to execute corresponding response measures.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the real-time intervention method based on blood glucose prediction for diabetic patients as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the real-time intervention method based on blood glucose prediction of a diabetic patient is implemented as described in any one of claims 1 to 7.

Images