Diabetic neuropathy dynamic early warning method based on multi-mode biological signals

Abstract

The invention provides a diabetic neuropathy dynamic early warning method based on a multi-modal biological signal, and the method comprises the steps: integrating a multi-modal sensor through an intelligent wearable device, and constructing a personalized prediction model in combination with a deep learning algorithm; through an artificial intelligence algorithm, diabetic neuropathy (DN) development risk prediction and unperceptual hypoglycemia real-time early warning are realized. According to the time-frequency domain characteristics and clinical data of the multi-mode signal, the time window of the DN from the progress to the serious stage can be predicted; recognizing an unperceptual hypoglycemia risk by combining ECG and HR features with a blood glucose trend prediction model, and triggering grading intervention measures; based on the individual physiological response mode of the patient, the algorithm threshold is dynamically optimized, and the accuracy of prediction and early warning is improved. According to the method, HR, BP, ECG and other multi-dimensional signals are integrated, and the prediction precision is improved; continuous monitoring and instant feedback are supported; the self-adaptive algorithm adapts to individual differences of patients, and the limitation of a traditional general model is overcome.

Description

Technical Field

[0001] The present invention relates to the cross - technical field of biomedical engineering and intelligent wearable devices, in particular to a dynamic early warning method for diabetic neuropathy based on multi - modal biological signals and a hypertension VR intervention device, an electronic device, and a computer - readable storage medium based on dynamic biofeedback and individualized modeling. Background Art

[0002] Diabetic neuropathy (DN), as one of the most common chronic complications of diabetes, its early diagnosis and disease course monitoring mainly rely on invasive examination methods such as nerve conduction tests and quantitative sensory tests. In current clinical practice, the early warning systems for asymptomatic hypoglycemia and the risk of neuropathy progression are mainly constructed based on the following technical frameworks: Single - dimensional biological signal monitoring: Using a continuous glucose monitoring system (CGMS) or a discrete electrocardiograph, risk judgment is carried out through a single glucose fluctuation curve or a heart rate variability (HRV) index with a fixed threshold. Such systems mostly rely on 30 - second ECG segment analysis and calculation of blood glucose standard deviation, lacking the time synchronization of multiple physiological dimensions.

[0003] Static risk assessment model: Using Logistic regression or traditional machine learning algorithms (such as random forest), a hierarchical prediction model is constructed based on quarterly clinical examination data (such as HbA1c, ankle - brachial index). The CAN risk score recommended by the American Diabetes Association (ADA) guidelines belongs to this type, and its parameter update cycle is as long as 3 - 6 months.

[0004] Artificial threshold alarm mechanism: Existing intelligent wearable devices trigger hypoglycemia early warnings by setting fixed thresholds (such as heart rate < 50 bpm for 5 minutes), and their parameter settings are based on population statistical characteristics rather than individualized physiological baselines.

[0005] However, the above - mentioned traditional existing technologies still have the following application defects: Risk of multi - modal data mismatch: Discrete monitoring devices result in time asynchrony (average deviation > 120 seconds) of signals such as ECG, blood pressure, and blood glucose. Especially in patients with autonomic nervous system dysfunction, the phenomenon that the cardiovascular compensatory response lags behind blood glucose changes (lag coefficient β = 0.32) will significantly reduce the prediction sensitivity.

[0006] Lack of dynamic pathological characterization: The time - domain features (such as SDNN) relied on by traditional models cannot capture the unique non - linear dynamic features of diabetic cardiac autonomic neuropathy. Research shows that when the sympathetic / parasympathetic nerve regulation ratio (LF / HF) > 3.5, the prediction accuracy of the FFT - based spectral analysis method for hypoglycemia drops by 37%.

[0007] Individual adaptability defect: The fixed threshold mechanism ignores the characteristics of autonomic nerve remodeling in DN patients (such as a 40% down-regulation of β2-adrenergic receptors), resulting in a decrease in specificity. Clinical data shows that the false alarm rate of the standard alarm threshold is as high as 68% in patients with a disease course > 10 years.

[0008] Artificial intervention lag: Existing systems require medical staff to manually calibrate model parameters (average cycle of 21 days) and cannot respond in real time to the non-linear physiological changes caused by acute metabolic disorders. In nocturnal asymptomatic hypoglycemia events, the delay in manual review results in an effective intervention rate of less than 42%.

[0009] Cross-modal association break: Current algorithms process blood glucose trends and cardiovascular signals separately and fail to establish a dynamic coupling model between the two. Especially when the baroreflex sensitivity (BRS) decreases to < 3 ms / mmHg due to autonomic neuropathy, the area under the ROC curve (AUC) of the single-modal prediction model decreases significantly by 0.15 - 0.22.

[0010] These defects result in a prediction error range of ±18 months for the risk of DN progression in existing technologies, a warning sensitivity of only 59% - 67% for unaware hypoglycemia, and a nocturnal false negative rate as high as 73% of the total system error. Especially the manual-dependent calibration mechanism makes the prediction model unable to adapt to the neuroplastic changes of DN patients in a timely manner, which has become the key bottleneck restricting accurate warning. Summary of the Invention

[0011] In order to solve the technical problems existing in the prior art, the present invention provides the following technical solutions: On the one hand, a dynamic warning method for diabetic neuropathy based on multi-modal biological signals is provided. This method is implemented by an electronic device and includes: S1. Real-time collect the biological signal data of DN patients: heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG), and input them into a preset preprocessing engine; S2. The preprocessing engine preprocesses the biological signal data and extracts and obtains the characteristic data of DN patients in real time: heart rate variability parameters (HRV), blood pressure variability parameters (BPV), and electrocardiogram morphology parameters (ECG-v); S3. Input the characteristic data into a pre-deployed DN dynamic warning model, and the model identifies the characteristic data and outputs the corresponding DN risk prediction results, including the probability of DN progression and the hypoglycemia risk index; S4. According to the risk prediction results, determine whether to trigger an alarm and output the corresponding judgment result; S5. Record the biological signal data, characteristic data, risk prediction results, and judgment results of DN patients, and store and upload them to the cloud.

[0012] Preferably, in step S1, the method for real-time collecting of the biological signal data of DN patients is as follows: Collect the biological signal data: heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG) through the intelligent wearable device worn by the DN patient; And, Use the motion artifact correction algorithm to correct the motion artifacts of the biological signal data, and upload the corrected data to the preprocessing engine.

[0013] Preferably, in step S3, the method for generating the DN early warning model includes the following steps: Obtain the biological signal data sets of a number of DN patients, including multimodal biological signals at different time series: heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG), and label data: DN severity grading and continuous glucose monitoring data (CGM); Preprocess the biological signal data of each DN patient through the preprocessing engine, and extract the corresponding feature data in real time: heart rate variability parameter (HRV), blood pressure variability parameter (BPV), and electrocardiogram morphology parameter (ECG-v); According to the DN severity grading, assign the corresponding DN progression probability label to the feature data; According to the continuous glucose monitoring data (CGM), assign the corresponding hypoglycemia risk index label to the feature data; Statistically analyze the feature data of each DN patient to obtain the data set for training the model, and divide the data set into a training set, a validation set, and a test set according to a ratio; Input the training set into a preset LSTM-CNN model for feature learning, and train to generate the initial DN early warning model; Use the validation set to verify the application performance of the DN early warning model: If the verification is passed, then use the test set to test the prediction performance of the DN early warning model: if passed, then deploy and apply the DN early warning model; otherwise, retrain; Otherwise, retrain.

[0014] Preferably, in step S4, according to the risk prediction result, determine whether to trigger an alarm and output the corresponding judgment result, including: Pre-set the warning thresholds for DN patients in terms of DN progression probability and hypoglycemia risk index respectively according to the individual data of DN patients, including: Based on the distribution differences of HRV / BPV / ECG-v between healthy people and DN patients, use kernel density estimation to determine the basic threshold; Based on the individual historical data of DN patients: the trend of HRV / BPV or the abnormal frequency of ECG-v, dynamically adjust the threshold using Bayesian optimization: , where: is the basic threshold; is the current threshold; is the optimized threshold; η is the set learning rate; is the set logarithmic likelihood gradient in the Bayesian function; Obtain the risk prediction results of DN patients, including the DN progression probability and hypoglycemia risk index of DN patients; Judge whether the DN progression probability of the DN patient exceeds the warning threshold of the DN patient in terms of the DN progression probability: If so, generate a DN clinical examination strategy with the corresponding probability; otherwise, abandon it; Judge whether the hypoglycemia risk index of the DN patient exceeds the warning threshold of the DN patient in terms of the hypoglycemia risk index: If so, trigger an alarm and generate an alarm signal for the corresponding DN patient; otherwise, abandon it.

[0015] On the other hand, a hypertension VR intervention device based on dynamic biofeedback and individualized modeling is provided. This device is applied to the dynamic early warning method of diabetic neuropathy based on multimodal bio-signals. The device includes: An intelligent wearable device for real-time collecting bio-signal data of DN patients: heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG), and inputting them into a preset preprocessing engine; A preprocessing engine for preprocessing the bio-signal data, and real-time extracting and obtaining the characteristic data of DN patients: heart rate variability parameters (HRV), blood pressure variability parameters (BPV), and electrocardiogram morphology parameters (ECG-v), and inputting the characteristic data into the early warning system; An early warning system for inputting the characteristic data into a pre-deployed DN dynamic early warning model, which identifies the characteristic data and outputs the corresponding DN risk prediction results, including DN progression probability and hypoglycemia risk index; A dynamic threshold module for judging whether the risk prediction results trigger an alarm according to the pre-set warning thresholds of DN patients respectively in terms of DN progression probability and hypoglycemia risk index, and outputting the corresponding judgment results; A data service module is used to record the bio-signal data, feature data, risk prediction results and judgment results of DN patients, and store and upload them to the cloud.

[0016] An intelligent wearable device (such as a smart watch) integrated with sensors of various modalities can be implemented and understood in combination with existing intelligent wearable devices, and can wirelessly transmit data to the background server.

[0017] On the other hand, an electronic device is provided, which includes: a processor; a memory, on which computer-readable instructions are stored, and when the computer-readable instructions are executed by the processor, any one of the methods in the above-mentioned dynamic early warning method for diabetic neuropathy based on multi-modal bio-signals is implemented.

[0018] On the other hand, a computer-readable storage medium is provided, in which at least one instruction is stored, and the at least one instruction is loaded and executed by a processor to implement any one of the methods in the above-mentioned dynamic early warning method for diabetic neuropathy based on multi-modal bio-signals.

[0019] The beneficial effects brought by the technical solutions provided in the embodiments of the present invention at least include: The present invention provides a dynamic early warning system for diabetic neuropathy based on multi-modal bio-signals. By integrating multi-modal sensors in an intelligent wearable device (such as a smart watch), signals such as heart rate (HR), blood pressure (BP) and single-lead electrocardiogram (ECG) are collected, and a personalized prediction model is constructed by combining deep learning algorithms. Through artificial intelligence algorithms, the risk prediction of the progression of diabetic neuropathy (DN) and the real-time early warning of hypoglycemia without awareness are realized. The following goals can be achieved: Dynamic risk stratification: According to the time-frequency domain characteristics of multi-modal signals and clinical data, predict the time window for DN to progress to the severe stage; Key index early warning: By combining the ECG and HR characteristics with the blood glucose trend prediction model, identify the risk of hypoglycemia without awareness and trigger graded intervention measures; Adaptive learning mechanism: Based on the individual physiological response patterns of patients, dynamically optimize the algorithm threshold to improve the accuracy of prediction and early warning.

[0020] The core advantages of the present invention are: Multi-modal fusion: Integrate multi-dimensional signals such as HR, BP and ECG to improve the prediction accuracy; Real-time: Support continuous monitoring and instant feedback; Personalization: Adapt to individual differences of patients through adaptive algorithms to overcome the limitations of traditional general models. Description of the Drawings

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0022] Figure 1 is a flowchart of a dynamic early warning method for diabetic neuropathy based on multimodal biological signals provided by an embodiment of the present invention; Figure 2 is a schematic diagram of the training mechanism of a DN early warning model provided by an embodiment of the present invention; Figure 3 is a block diagram of a hypertension VR intervention device based on dynamic biofeedback and individualized modeling provided by an embodiment of the present invention; Figure 4 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0023] The following describes the technical solutions in the present invention with reference to the accompanying drawings.

[0024] In the embodiments of the present invention, words such as "exemplarily" and "for example" are used to represent examples, illustrations or explanations. Any embodiment or design solution described as an "example" in the present invention should not be construed as being more preferred or having more advantages than other embodiments or design solutions. Exactly speaking, the use of the word "example" is intended to present concepts in a specific way. In addition, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or either of them can be selected.

[0025] In the embodiments of the present invention, "image" and "picture" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same. "(of)", "corresponding", and "corresponding" can sometimes be used interchangeably. It should be noted that when the difference is not emphasized, the meanings they express are the same.

[0026] In the embodiments of the present invention, sometimes subscripts such as W1 may be miswritten as non-subscript forms such as W1. When the difference is not emphasized, the meanings they express are the same.

[0027] To make the technical problems, technical solutions and advantages to be solved by the present invention clearer, the following will be described in detail with reference to the accompanying drawings and specific embodiments.

[0028] An embodiment of the present invention provides a dynamic early warning method for diabetic neuropathy based on multimodal biological signals. This method can be implemented by an electronic device, which can be a terminal or a server. As Figure 1 shown in the flowchart of the dynamic early warning method for diabetic neuropathy based on multimodal biological signals, the processing flow of this method can include the following steps: S1. Real-time collect the biological signal data of DN patients: heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG), and input them into a preset preprocessing engine; S2. The preprocessing engine preprocesses the biological signal data, and real-time extracts and obtains the characteristic data of DN patients: heart rate variability parameters (HRV), blood pressure variability parameters (BPV), and electrocardiogram morphology parameters (ECG-v); S3. Input the characteristic data into a pre-deployed DN dynamic early warning model, and the model identifies the characteristic data and outputs the corresponding DN risk prediction results, including the DN progression probability and the hypoglycemia risk index; S4. According to the risk prediction results, determine whether to trigger an alarm, and output the corresponding judgment results; S5. Record the biological signal data, characteristic data, risk prediction results, and judgment results of DN patients, and store and upload them to the cloud.

[0029] Combined with the Figure 3 shown device structure schematic diagram. The implementation of the present invention mainly integrates multimodal sensors through intelligent wearable devices (such as smart watches) to collect signals such as heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG), and constructs a personalized prediction model in combination with deep learning algorithms to predict the DN progression probability and hypoglycemia risk index of DN patients. Further, through the combination of ECG and HR features and the blood glucose trend prediction model, the risk of hypoglycemia unawareness is identified, and graded intervention measures are triggered.

[0030] At the same time, the system has an adaptive learning mechanism: based on the individual physiological response pattern of the patient, the algorithm threshold is dynamically optimized to improve the accuracy of prediction and early warning.

[0031] Hospitals can distribute intelligent wearable devices to DN patients, file them in the background, and bind information such as device ID and patient medical record number. After the device is activated, the data reported by the device can be written into the patient's electronic medical record file. The background is deployed with a preprocessing engine, an early warning system, a dynamic threshold module, a system output module, etc. (specifically understood in combination with the device in this application), which can realize corresponding data analysis, intelligent analysis calculation, and judgment with the support of the computing power of the CPU processing system.

[0032] The following will be described in detail.

[0033] Preferably, in step S1, the method for collecting the biological signal data of DN patients in real time is as follows: Collect the biological signal data, namely heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG), through the intelligent wearable device worn by the DN patient; And Use the motion artifact correction algorithm to correct the motion artifacts of the biological signal data, and upload the corrected data to the preprocessing engine.

[0034] The intelligent wearable device integrates multiple sensors, a power supply, an MCU chip, a memory, a processor, a wireless module, etc. (specifically refer to existing intelligent bracelets, etc.), and can collect multi-modal biological signals such as HR, BP, and ECG.

[0035] The intelligent wearable device integrates multi-modal sensors: for example, it uses a photoplethysmography (PPG) sensor to measure HR, the oscillometric method to measure BP, and the sampling rate is ≥200 Hz; it integrates a single-lead ECG electrode to collect electrocardiogram signals (it can also integrate an accelerometer, etc. for motion monitoring and correction); the signal is preprocessed by the processor, such as A / D conversion, etc., and through the built-in motion artifact correction algorithm, the stability of the signal in a dynamic environment is ensured.

[0036] The motion artifact correction algorithm can correct the motion artifacts of the sampled signal, improve the accuracy of the signal features, and facilitate the model to better identify the features. For example, to solve the problems of PPG waveform distortion (such as peak inversion) and ECG baseline drift caused by motion artifacts in traditional methods, the following motion artifact correction algorithm is adopted: 1. Multi-modal time synchronization correction: , is the multi-modal signal (including PPG, ECG, BP) after time synchronization; S i is the i-th modal signal; is the inherent delay of the i-th sensor (for example, PPG: 50 ms, ECG: 20 ms, BP: 100 ms); is the Dirac impulse function, used to align the timestamps; is the dynamic compensation factor (calculated based on the accelerometer data).

[0037] 2. Motion noise separation and correction: , is the motion artifact component; is a three-axis accelerometer signal (sampling rate ≥ 200 Hz); is a wavelet packet decomposition weight matrix (Daubechies-4 basis function), where, uses 4-layer decomposition to extract the 0.5 - 10 Hz frequency band (corresponding to the human motion frequency), and the formula is; , is the j-th layer wavelet basis function, is the energy normalization coefficient (C1 = 0.6, C2 = 0.3, C3 = 0.08, C4 = 0.02); is a convolution operation used to extract frequency domain coupling features; is independent component analysis to separate the signal and noise sources.

[0038] 3. Adaptive filtering compensation: , is the corrected output signal; α is the dynamic attenuation coefficient (range 0.2 - 1.0, controlled by the signal signal-to-noise ratio SNR), where: , k is an empirical constant, preferably taking the value: k = 0.5, SNR is calculated by the ratio of the AC / DC components of the PPG signal, and the filtering intensity is dynamically adjusted; is a non-linear activation function to suppress low-frequency motion interference.

[0039] Through multi-modal signal fusion and dynamic parameter adjustment, it solves the problems of PPG waveform distortion (such as peak inversion) and ECG baseline drift caused by motion artifacts in traditional methods, and reduces the measurement error of signals such as heart rate.

[0040] The signal is uploaded to the background, and the preprocessing engine in the background performs preprocessing and extracts signal features.

[0041] Functions of the preprocessing engine: Clean and extract features from the original signal. Perform real-time band-pass filtering (0.5 - 40 Hz) to remove noise and baseline drift. Extract feature extraction, such as: The R-wave detection algorithm extracts the QRS complex wave and calculates the heart rate variability (HRV); The wavelet packet transform calculates the multi-scale entropy of HRV and extracts time-frequency domain features. Specifically:‌ 1. Core process of the preprocessing engine‌ Signal denoising and baseline correction‌ Adaptive filtering: The Kalman filter is combined with wavelet domain decomposition to eliminate motion artifacts and high-frequency noise. For ECG signals, a 0.05 - 40 Hz band-pass filter is selected to suppress electromyogram interference and power frequency noise.

[0042] Baseline drift correction: The low-frequency baseline drift of the ECG signal is eliminated through cubic spline interpolation to ensure the stability of QT interval measurement.

[0043] Multi-modal time synchronization: The accelerometer data is used to compensate for the time deviation (±50 ms) between PPG and ECG to ensure the time alignment of HRV and BPV parameters.

[0044] Signal segmentation and sliding window mechanism Dynamic window division: The window length (default 2 seconds) is dynamically adjusted based on the R-wave interval to adapt to the signal cycle characteristics at different heart rates.

[0045] Overlapping sampling strategy: The window overlap rate is 50% to improve the calculation continuity of short-term HRV parameters (such as SDNN).

[0046] 2. Heart rate variability (HRV) parameter extraction R-wave detection and QRS complex localization Differential threshold algorithm: Combining the first-order difference and adaptive threshold to detect the R-wave peak, with better anti-interference ability than the traditional Pan-Tompkins algorithm.

[0047] Morphological correction: Identifying the start and end points of QRS through the second derivative method to eliminate false detections caused by tall T waves.

[0048] Multi-scale entropy and time-frequency domain feature calculation Wavelet packet decomposition: Using the Daubechies-4 wavelet basis for 5-layer decomposition to extract the energy ratio in the frequency range of 0.04 - 0.4 Hz (corresponding to the sympathetic / parasympathetic nerve activity frequency band).

[0049] Nonlinear dynamics parameters: Calculating sample entropy and multifractal spectrum to quantify the chaotic characteristics of autonomic regulation.

[0050] Typical HRV parameters Time domain parameters: SDNN (standard deviation), RMSSD (root mean square difference of adjacent RR intervals), PNN50 (proportion of adjacent RR interval differences > 50 ms).

[0051] Frequency domain parameters: LF / HF ratio (low-frequency / high-frequency energy ratio, reflecting sympathetic-parasympathetic balance), total power (TP).

[0052] 3. Extraction of Blood Pressure Variability (BPV) Parameters Preprocessing of Blood Pressure Time Series Removal of Trend Terms: Separating the slow-changing trend and fast-fluctuating components of blood pressure through moving average filtering (window length 30 seconds).

[0053] Correction of Outliers: Eliminating blood pressure spikes caused by motion artifacts based on median filtering (window length 5 samplings).

[0054] BPV Feature Quantification Method Time-Frequency Analysis: Calculating the energy density distribution of blood pressure fluctuations in the frequency band of 0.01 - 0.15 Hz using continuous wavelet transform (CWT).

[0055] Statistical Features: Blood Pressure Standard Deviation (BP-SD), Coefficient of Variation (CV = SD / Mean × 100%), Dynamic Trend Slope (DTS).

[0056] 4. Extraction of Electrocardiogram Morphological Parameters (ECG-v) Key Waveform Recognition Algorithm QT Interval Measurement: Locating the end point of the T wave based on the dynamic threshold method and eliminating the influence of heart rate using the heart rate correction formula.

[0057] ST Segment Analysis: Sampling the slope of the ST segment 80 ms after the J point and quantifying ischemic ST segment depression / elevation through polynomial fitting.

[0058] T Wave Symmetry Index: Calculating the time ratio of the ascending and descending branches of the T wave, with the normal range being 0.88 - 1.12.

[0059] Morphological Dynamic Tracking Dynamic Time Warping (DTW): Comparing the morphological differences between the current heartbeat and the template waveform to quantify myocardial repolarization dispersion.

[0060] Covariance Matrix Analysis: Constructing the covariance matrix features of the QRS-T wave complex to detect minor repolarization abnormalities.

[0061] 5. Optimization of Real-Time Computing Architecture Heterogeneous Parallel Processing: Allocating ECG feature extraction and HRV calculation to the FPGA hardware acceleration module, and performing BPV analysis by the ARM Cortex-M4 core, reducing the latency to 28 ms.

[0062] Memory Optimization Strategy: Using a ring buffer to store sliding window data, reducing the memory occupancy by 40%.

[0063] Model Lightweighting: Compressing the number of parameters of the GRU network to 50 KB, enabling real-time operation on wearable device terminals.

[0064] By integrating time-frequency analysis (such as wavelet packet transform), nonlinear dynamics (such as multi-scale entropy), and morphological quantification (such as dynamic tracking of QT interval), the accuracy of autonomic nerve function assessment for patients with diabetic neuropathy (DN) has been significantly improved.

[0065] Such as Figure 2 As shown, preferably, in step S3, the method for generating the DN early warning model includes the following steps: Obtain a biological signal dataset of several DN patients, including multi-modal biological signals at different time sequences: heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG), as well as label data: DN severity grading and continuous glucose monitoring data (CGM); Preprocess the biological signal data of each DN patient through the preprocessing engine, and extract corresponding feature data in real time: heart rate variability parameters (HRV), blood pressure variability parameters (BPV), and electrocardiogram morphology parameters (ECG-v); According to the DN severity grading, assign corresponding DN progression probability labels to the feature data; According to the continuous glucose monitoring data (CGM), assign corresponding hypoglycemia risk index labels to the feature data; Statistically analyze the feature data of each DN patient to obtain a dataset for training the model, and divide the dataset into a training set, a validation set, and a test set according to a ratio; Input the training set into a preset LSTM-CNN model for feature learning, and train to generate the initial DN early warning model; Use the validation set to verify the application performance of the DN early warning model: If the verification is passed, use the test set to test the prediction performance of the DN early warning model: if passed, deploy and apply the DN early warning model; otherwise, retrain; Otherwise, retrain.

[0066] Function of the early warning system (calculating and analyzing the early warning results based on the deployed algorithm chip): predicting DN progression and hypoglycemia risk based on multi-modal signals. Adopt a hybrid model combining long short-term memory network (LSTM) and convolutional neural network (CNN): Input parameters: time series data of standardized HRV, blood pressure variability (BPV), and ECG morphology parameters (such as QT interval, ST segment changes); Output results: DN progression probability (0-1) and hypoglycemia risk index (grading: low, medium, high).

[0067] Specific model training and application are as follows: 1. Data Preparation Input Data: Obtain the pre - processed and feature - extracted time - series data from the pre - processing engine, including: Heart Rate Variability (HRV) parameters: Such as multi - scale entropy, frequency - domain features (LF / HF ratio, etc.).

[0068] Blood Pressure Variability (BPV) parameters: Such as the time - series of blood pressure fluctuations.

[0069] Electrocardiogram Morphological Parameters (ECG - v): Such as QT interval, ST - segment changes, T - wave symmetry, etc.

[0070] Data Standardization: Standardize different features to ensure that the numerical ranges of each feature are consistent, avoiding biases caused by dimensional differences during model training. Common methods include: Z - score Normalization: Convert the data into a distribution with a mean of 0 and a standard deviation of 1.

[0071] Min - Max Scaling: Scale the data to the interval [0, 1].[[]END]

[0072] 2. Model Architecture Design Model Selection: Adopt a hybrid model that combines Long Short - Term Memory Network (LSTM) and Convolutional Neural Network (CNN): LSTM: Good at processing time - series data and capturing the long - term dependencies of signals such as HRV and BPV changing over time.

[0073] CNN: Used to extract the spatial features of ECG signals, such as local morphological patterns of waveforms.

[0074] Model Structure: Input Layer: Receive the standardized multi - modal time - series data, and the data dimension is (number of samples, time steps, number of features).

[0075] CNN Layer: Apply a one - dimensional convolutional layer to the ECG signal, and use multiple convolutional kernels to extract local features (such as QRS complex, T - wave morphology).

[0076] LSTM Layer: Input the features extracted by CNN and the HRV and BPV time - series data into the LSTM layer to learn the changing trends of signals over time.

[0077] Fusion Layer: Fusion the outputs of CNN and LSTM through a fully - connected layer to generate a comprehensive feature (F) representation: , The training weight matrix is used to linearly weight the CNN output feature C and the LSTM output feature L. WC measures the importance of the spatial features extracted by the CNN (such as ECG waveform abnormalities) for the fused feature, WL measures the contribution of the temporal features extracted by the LSTM (such as HRV circadian rhythm) to the fused feature, b (bias term) provides a baseline offset for the fused feature, and σ (ReLU activation function).

[0078] Output layer: DN progression prediction: Outputs a probability value between 0 and 1, indicating the probability of DN progressing to the severe stage in a future period of time.

[0079] Hypoglycemia risk prediction: Outputs a graded risk index (e.g., 0: low risk, 1: medium risk, 2: high risk).

[0080] 3. Model training Dataset: Use a dataset containing a large number of biometric signal records of diabetic patients, which is required to include: Temporal data of multimodal biometric signals.

[0081] Label data: DN severity grading (mild, moderate, severe) and continuous glucose monitoring (CGM) data Training process: Divide the dataset into a training set, a validation set, and a test set, and use cross-validation to optimize the model hyperparameters.

[0082] Train the model using the backpropagation algorithm and the Adam optimizer. The loss functions available are: Cross-entropy loss: Suitable for classification tasks (such as hypoglycemia risk grading).

[0083] Mean squared error: Suitable for regression tasks (such as DN progression probability prediction).

[0084] Monitor the performance of the validation set to prevent overfitting.

[0085] Model evaluation: Evaluate the prediction accuracy, sensitivity, and specificity of the model on the test set to ensure the generalization ability.

[0086] On the test set, F1-score is used for DN prediction (considering class imbalance), and mean absolute error (MAE) is used for hypoglycemia warning.

[0087] For the dynamic threshold performance, evaluate different threshold strategies through the area under the receiver operating characteristic curve (ROC). It is required that the AUC of the personalized threshold is at least 5% higher than that of the static threshold.

[0088] If the model validation test passes, it can be deployed on the backend system for DN early warning prediction.

[0089] 4. Risk prediction Real-time application: The core algorithm layer receives the bio-signal data that has been real-time collected and processed.

[0090] Forward propagation: Input the data into the trained model, and the output is: DN progression probability: A value between 0 and 1, representing the probability that DN will progress to the severe stage within the next 3 / 6 / 12 months.

[0091] Hypoglycemia risk index: A graded value (e.g., 0: low risk, 1: medium risk, 2: high risk), based on the current signal characteristics and historical trends.

[0092] Output application: When the DN progression probability exceeds 0.5, more frequent clinical examinations are recommended.

[0093] When the hypoglycemia risk index is 2 (high risk), the system triggers an alarm.

[0094] 5. Model update and optimization Online learning mechanism: Dynamically update the model according to the patient's individual data to adapt to physiological changes.

[0095] Update strategy: Regularly (e.g., monthly) fine-tune the model using the patient's recent data and clinical feedback.

[0096] Adopt transfer learning and only update some layers to maintain computational efficiency.

[0097] Threshold adjustment: Combine a dynamic threshold module to adjust the alarm trigger threshold according to the prediction results and the patient's historical data.

[0098] The early warning system is capable of processing multi-modal bio-signal data obtained from the signal acquisition layer and the preprocessing engine, and using a deep learning model to predict the risk of diabetic neuropathy (DN) progression and the risk of hypoglycemia unawareness.

[0099] Preferably, in step S4, according to the risk prediction result, determine whether to trigger an alarm and output the corresponding judgment result, including: Pre-set the early warning thresholds for DN patients in terms of DN progression probability and hypoglycemia risk index respectively based on the individual data of DN patients, including: Based on the differences in HRV / BPV / ECG-v distributions between healthy people and DN patients, use kernel density estimation to determine the basic threshold; Through the individual historical data of DN patients: the trend of HRV / BPV or the abnormal frequency of ECG-v, use Bayesian optimization to dynamically adjust the threshold: , Wherein: is the base threshold; is the current threshold; is the optimized threshold; η is the set learning rate; is the set log-likelihood gradient in the Bayesian function; Obtain the risk prediction result of the DN patient, including the DN progression probability and the hypoglycemia risk index of the DN patient; Judge whether the DN progression probability of the DN patient exceeds the warning threshold of the DN patient in terms of the DN progression probability: If so, generate a DN clinical examination strategy with the corresponding probability; otherwise, give up; Judge whether the hypoglycemia risk index of the DN patient exceeds the warning threshold of the DN patient in terms of the hypoglycemia risk index: If so, trigger an alarm and generate an alarm signal for the corresponding DN patient; otherwise, give up.

[0100] Because each DN patient has individual differences, different warning thresholds need to be set for them for individualized warning. Here, a dynamic threshold module is set up in the background, which can dynamically adjust its warning threshold based on the individual data of each DN patient, specifically including the warning thresholds for the DN progression probability and the hypoglycemia risk index respectively.

[0101] Based on the differences in HRV / BPV / ECG-v distributions between healthy people and DN patients, kernel density estimation is used to determine the base threshold. For example, taking the HRV distribution difference as an example: Based on the HRV parameters of healthy people and DN patients (such as SDNN, RMSSD, LF / HF ratio), construct a non-parametric probability density function: , is the HRV sample data of the DN patient, and x is the HRV sample data of healthy people (for example, there are n≥1000 cases of healthy people and n≥300 cases of DN patients); is the Gaussian kernel function, and the bandwidth h is optimized by the Silverman criterion; Principle for delimiting the base threshold: Take the 95th percentile of the healthy population distribution as the initial abnormal threshold (for example, SDNN < 50ms triggers a warning).

[0102] For the steps of Bayesian optimization, please combine with the Bayesian optimization framework: Gaussian Process Regression (GP): Construct the mapping relationship between the threshold parameters and the warning performance for understanding.

[0103] By integrating population distribution characteristics and individual dynamic data, the adaptive defects of traditional fixed-threshold models in diabetic neuropathy monitoring are overcome, which is particularly suitable for the dynamic tracking of non-linear attenuation of HRV (such as LF / HF ratio mutation) and ECG repolarization abnormalities caused by autonomic nerve remodeling.

[0104] When this system is applied: The user wears and activates the smart wearable device, and the device collects continuous data and uploads it to the background; The preprocessing engine extracts features in real time (such as HRV multi-scale entropy, ECG morphological parameters); The feature data is input into the trained LSTM-CNN model (DN early warning model) to generate risk prediction results; The dynamic threshold module evaluates the risk index and decides whether to trigger an alarm (such as vibration prompt or push notification to the PDA / EDA nursing terminal); The system stores the results and uploads them to the cloud regularly for medical staff to analyze (the administrator can view them in the background).

[0105] Figure 3 It is a block diagram of a hypertension VR intervention device based on dynamic biofeedback and individualized modeling shown according to an exemplary embodiment. This device is used for the dynamic early warning method of diabetic neuropathy based on multi-modal bio-signals. Refer to Figure 3 On the other hand, a hypertension VR intervention device based on dynamic biofeedback and individualized modeling is provided. This device is applied to the dynamic early warning method of diabetic neuropathy based on multi-modal bio-signals. The device includes: A smart wearable device for real-time collecting bio-signal data of DN patients: heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG), and inputting them into a preset preprocessing engine; A preprocessing engine for preprocessing the bio-signal data, and real-time extracting and obtaining feature data of DN patients: heart rate variability parameters (HRV), blood pressure variability parameters (BPV), and electrocardiogram morphological parameters (ECG-v), and inputting the feature data into the early warning system; An early warning system for inputting the feature data into a pre-deployed DN dynamic early warning model, which identifies the feature data by this model and outputs corresponding DN risk prediction results, including DN progression probability and hypoglycemia risk index; A dynamic threshold module for judging whether the risk prediction result triggers an alarm according to the pre-set early warning thresholds of DN patients respectively on the DN progression probability and hypoglycemia risk index, and outputting corresponding judgment results; A data service module for recording the biosignal data, feature data, risk prediction results, and judgment results of DN patients, and storing and uploading them to the cloud.

[0106] Figure 4 FIG. 4 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. As Figure 4 shown, the electronic device may include the above-mentioned Figure 3 hypertensive VR intervention device based on dynamic biofeedback and individualized modeling shown in FIG. Optionally, the electronic device 410 may include a first processor 2001.

[0107] Optionally, the electronic device 410 may further include a memory 2002 and a transceiver 2003.

[0108] Among them, the first processor 2001 is connected to the memory 2002 and the transceiver 2003, such as through a communication bus.

[0109] Next, each component of the electronic device 410 will be specifically introduced in conjunction with Figure 4 FIG. Among them, the first processor 2001 is the control center of the electronic device 410, which may be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 is one or more central processing units (CPUs), or may be a specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention, such as: one or more digital signal processors (DSPs), or one or more field programmable gate arrays (FPGAs).

[0110] Optionally, the first processor 2001 may execute various functions of the electronic device 410 by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0111] In a specific implementation, as an embodiment, the first processor 2001 may include one or more CPUs, such as Figure 4 the CPU0 and CPU1 shown in FIG.

[0112] In a specific implementation, as an embodiment, the electronic device 410 may also include multiple processors, such as Figure 4The first processor 2001 and the second processor 2004 shown in []. Each of these processors can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). Here, the processor can refer to one or more devices, circuits, and / or processing cores for processing data (such as computer program instructions).

[0113] Among them, the memory 2002 is used to store the software program for implementing the solution of the present invention and is controlled by the first processor 2001 for execution. The specific implementation method can refer to the above method embodiment and will not be elaborated here.

[0114] Optionally, the memory 2002 can be a read-only memory (ROM) or other types of static storage devices that can store static information and instructions, a random access memory (RAM) or other types of dynamic storage devices that can store information and instructions, or can also be an electrically erasable programmable read-only memory (EEPROM), a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto. The memory 2002 can be integrated with the first processor 2001 or exist independently and is coupled to the first processor 2001 through the interface circuit of the electronic device 410 ( Figure 4 not shown in []). The embodiments of the present invention do not make specific limitations on this.

[0115] The transceiver 2003 is used to communicate with a network device or communicate with a terminal device.

[0116] Optionally, the transceiver 2003 can include a receiver and a transmitter ( Figure 4 not shown separately in []). Among them, the receiver is used to implement the receiving function, and the transmitter is used to implement the sending function.

[0117] Optionally, the transceiver 2003 can be integrated with the first processor 2001 or exist independently and is coupled to the first processor 2001 through the interface circuit of the electronic device 410 ( Figure 4 not shown in []). The embodiments of the present invention do not make specific limitations on this.

[0118] It should be noted thatFigure 4 The structure of the electronic device 410 shown does not constitute a limitation on the router. The actual knowledge structure recognition device may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0119] In addition, for the technical effects of the electronic device 410, reference may be made to the technical effects of the method for dynamically warning of diabetic neuropathy based on multimodal biosignals described in the foregoing method embodiments, which will not be elaborated herein.

[0120] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0121] It should also be understood that the memory in the embodiments of the present invention can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchlink DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0122] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present invention are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center by wired (such as infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or a data center that contains one or more collections of available media. The available media can be magnetic media (such as floppy disks, hard disks, magnetic tapes), optical media (such as DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0123] It should be understood that the term "and / or" in this document is merely a description of the association relationship between associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist simultaneously, and B exists alone. Here, A and B can be singular or plural. In addition, the character " / " in this document generally represents an "or" relationship between the associated objects before and after, but it may also represent an "and / or" relationship, which can be specifically understood by referring to the context before and after.

[0124] In the present invention, "at least one" means one or more, and "a plurality" means two or more. "At least one of the following" or its similar expressions refer to any combination of these items, including any combination of single items or plural items. For example, at least one of a, b, or c can represent: a, b, c, a - b, a - c, b - c, or a - b - c, where a, b, and c can be single or multiple.

[0125] It should be understood that in various embodiments of the present invention, the magnitudes of the sequence numbers of the above processes do not mean the order of execution. The order of execution of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0126] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or by a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Skilled professionals can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the present invention.

[0127] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the devices, apparatuses, and units described above can refer to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0128] In several embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0129] 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 can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0130] In addition, the functional units in each embodiment of the present invention can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.

[0131] When the above-mentioned function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0132] The above is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.

Claims

1. A dynamic early warning method for diabetic neuropathy based on multimodal biological signals, characterized in that, The method includes the following steps: S1. Real-time collect the biological signal data of DN patients: heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG), and input them into a preset preprocessing engine; S2. The preprocessing engine preprocesses the biological signal data, and extracts and obtains the characteristic data of DN patients in real time: heart rate variability parameters (HRV), blood pressure variability parameters (BPV), and electrocardiogram morphology parameters (ECG-v); S3. Input the characteristic data into a pre-deployed DN dynamic warning model, which identifies the characteristic data and outputs the corresponding DN risk prediction results, including the DN progression probability and the hypoglycemia risk index; S4. According to the risk prediction results, determine whether to trigger an alarm and output the corresponding judgment result; S5. Record the biological signal data, characteristic data, risk prediction results, and judgment results of DN patients, and store and upload them to the cloud.

2. The dynamic early warning method for diabetic neuropathy based on multimodal biological signals according to claim 1, wherein In step S1, the method for real-time collecting the biological signal data of DN patients is as follows: Collect the biological signal data: heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG) through an intelligent wearable device worn by DN patients; And Use a motion artifact correction algorithm to correct the motion artifacts of the biological signal data, and upload the corrected data to the preprocessing engine.

3. The dynamic early warning method for diabetic neuropathy based on multimodal biological signals according to claim 1, wherein In step S3, the method for generating the DN warning model includes the following steps: Obtain the biological signal data sets of a number of DN patients, including multi-modal biological signals at different time series: heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG), and label data: DN severity grading and continuous glucose monitoring data (CGM); The preprocessing engine preprocesses the biological signal data of each DN patient, and extracts the corresponding characteristic data in real time: heart rate variability parameters (HRV), blood pressure variability parameters (BPV), and electrocardiogram morphology parameters (ECG-v); According to the DN severity grading, assign corresponding DN progression probability labels to the characteristic data; According to the continuous glucose monitoring data (CGM), assign corresponding hypoglycemia risk index labels to the characteristic data; Statistically analyze the characteristic data of each DN patient to obtain a data set for training the model, and divide the data set into a training set, a validation set, and a test set according to a ratio; Input the training set into a preset LSTM-CNN model for feature learning, and train to generate the initial DN warning model; Use the validation set to verify the application performance of the DN warning model: If the verification is passed, use the test set to test the prediction performance of the DN warning model: if it passes, deploy and apply the DN warning model; otherwise, retrain; Otherwise, retrain.

4. The dynamic early warning method for diabetic neuropathy based on multimodal biological signals according to claim 1, wherein In step S4, according to the risk prediction results, determine whether to trigger an alarm and output the corresponding judgment result, including: Pre-set the warning thresholds for DN patients in terms of DN progression probability and hypoglycemia risk index respectively according to the individual data of DN patients, including: Based on the distribution differences of HRV / BPV / ECG-v between healthy people and DN patients, the base threshold is determined using kernel density estimation; Through the individual historical data of DN patients: the trend of HRV / BPV or the abnormal frequency of ECG-v, the threshold is dynamically adjusted using Bayesian optimization: ， Where: as the base threshold; is the current threshold value; is the optimized threshold value; η is the set learning rate; The logarithmic likelihood gradient set in the Bayesian function; Obtain the risk prediction results of DN patients, including the DN progression probability and hypoglycemia risk index of DN patients; Judge whether the DN progression probability of DN patients exceeds the warning threshold of DN patients in terms of DN progression probability: If so, generate a DN clinical examination strategy with the corresponding probability; otherwise, give up; Judge whether the hypoglycemia risk index of DN patients exceeds the warning threshold of DN patients in terms of hypoglycemia risk index: If so, trigger an alarm and generate an alarm signal for the corresponding DN patient; otherwise, give up.

5. A hypertension VR intervention device based on dynamic biofeedback and individualized modeling, the hypertension VR intervention device based on dynamic biofeedback and individualized modeling is used to implement the dynamic early warning method for diabetic neuropathy based on multimodal bio-signals as described in any one of claims 1-4, characterized in that, The device includes: An intelligent wearable device for real-time collection of biometric signal data of DN patients: heart rate (HR), blood pressure (BP), and single-lead electrocardiogram (ECG), and inputting them into a preset preprocessing engine; A preprocessing engine for preprocessing the biometric signal data, and real-time extraction and acquisition of characteristic data of DN patients: heart rate variability parameters (HRV), blood pressure variability parameters (BPV), and electrocardiogram morphology parameters (ECG-v), and inputting the characteristic data into an early warning system; An early warning system for inputting the characteristic data into a pre-deployed DN dynamic early warning model, which identifies the characteristic data and outputs the corresponding DN risk prediction results, including DN progression probability and hypoglycemia risk index; A dynamic threshold module for judging whether the risk prediction results trigger an alarm according to the pre-set warning thresholds of DN patients in terms of DN progression probability and hypoglycemia risk index respectively, and outputting the corresponding judgment results; A data service module for recording the biometric signal data, characteristic data, risk prediction results, and judgment results of DN patients, and storing and uploading them to the cloud.

6. An electronic device, characterized in that, The electronic device includes: A processor; A memory, on which computer-readable instructions are stored, and when the computer-readable instructions are executed by the processor, the method described in any one of claims 1 to 4 is implemented.

7. A computer-readable storage medium, characterized in that, Program code is stored in the computer-readable storage medium, and the program code can be called by the processor to execute the method described in any one of claims 1 to 4.

Images