Coronary heart disease risk prediction system based on remote traditional Chinese medicine diagnosis and treatment

Abstract

The invention, which relates to the technical field of clinical medicine, discloses a coronary heart disease risk prediction system based on remote traditional Chinese medicine diagnosis and treatment, comprising the following modules: a data acquisition module, a data preprocessing module, a signal enhancement module, a data transmission module, a machine learning analysis module, an expert evaluation module and a feedback module. The data acquisition module is used for acquiring an electrocardiosignal of a patient through wearable electrocardiosignal acquisition equipment. According to the coronary heart disease risk prediction system based on remote traditional Chinese medicine diagnosis and treatment, acquired signals can be enhanced through the signal enhancement module, real features of electrocardiosignals are recovered, and a patient is prompted to adjust the position of an electrode plate or change the acquisition posture through voice and characters, so that operation of the patient is further standardized; the problem that the accuracy of coronary heart disease risk prediction is affected due to the fact that collected electrocardio data has a large amount of noise and artifacts is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of clinical medicine, and in particular to a coronary heart disease risk prediction system based on remote traditional Chinese medicine diagnosis and treatment. Background Art

[0002] Coronary heart disease, also known as coronary artery disease, is a heart disease that is mainly caused by the stenosis or blockage of the coronary arteries, resulting in insufficient blood supply to the myocardium. The coronary arteries are blood vessels that supply oxygen and nutrients to the heart muscles. The main cause of coronary heart disease is coronary atherosclerosis, which is the formation of plaques on the inner walls of the coronary arteries. These plaques are composed of fat, cholesterol, calcium and other substances, which cause the blood vessels to narrow and reduce or block blood flow. Coronary heart disease is one of the leading causes of death worldwide. Traditional methods for predicting the risk of coronary heart disease mainly rely on statistical models, such as the Framingham risk score, which are usually based on Western medical indicators such as the patient's age, gender, blood pressure, cholesterol level, history of diabetes and smoking habits.

[0003] At present, a coronary heart disease risk prediction system based on remote Chinese medicine diagnosis and treatment has been used to predict the risk of coronary heart disease. Although it can predict the patient's risk of coronary heart disease or determine whether he or she has coronary heart disease, in the remote Chinese medicine diagnosis and treatment coronary heart disease risk prediction system, although patients can use wearable devices or simple ECG acquisition devices to collect ECG data at home, the collected ECG data often contain a lot of noise and artifacts due to improper operation by patients, such as inaccurate electrode placement, body shaking, or equipment being subject to surrounding electromagnetic interference, which affects the accuracy of coronary heart disease risk prediction. Summary of the invention

[0004] In order to solve the above technical problems, the present invention is implemented through the following technical solutions: a coronary heart disease risk prediction system based on remote traditional Chinese medicine diagnosis and treatment, comprising the following modules: a data acquisition module, a data preprocessing module, a signal enhancement module, a data transmission module, a machine learning analysis module, an expert evaluation module and a feedback module; The data acquisition module is used to acquire the patient's ECG signal through a wearable ECG acquisition device, and convert the analog signal into a digital signal through an analog-to-digital converter to obtain ECG data. The ECG acquisition device is equipped with a remote communication function to achieve real-time data transmission and ensure the stability and accuracy of the acquisition process. The data preprocessing module is used to filter and denoise the collected ECG data to eliminate interference signals and ensure data quality; the signal enhancement module is used to enhance the preprocessed ECG signal through an adaptive filtering algorithm to improve the clarity and stability of the signal; the data transmission module is used to upload the enhanced ECG data to the server via a wireless network, supporting multiple wireless communication protocols, including Bluetooth, Wi-Fi and cellular mobile networks, to ensure smooth data transmission; The machine learning analysis module analyzes the enhanced ECG data based on a convolutional neural network to predict the patient's risk of coronary heart disease and obtain a prediction result; The expert evaluation module is used to perform expert-level evaluation based on the prediction results of the machine learning analysis module and in combination with the patient's TCM constitution characteristics and medical history data to obtain evaluation results and provide personalized health advice based on the nine constitution types; The feedback module feeds back the evaluation results to the patient and the doctor through voice prompts and a visual interface, and provides corresponding health advice.

[0005] The process of obtaining the assessment results and providing personalized health advice based on the nine physical types is as follows: The prediction results of the patient's coronary heart disease risk are obtained from the machine learning analysis module. At the same time, the TCM constitution characteristic data of the patient are collected through the TCM constitution classification questionnaire. The TCM constitution classification questionnaire covers the patient's dietary preferences, sleep conditions, daily activities, and emotional state, and is used to determine which one or more of the patient's constitutions are mild, qi-deficient, yang-deficient, yin-deficient, phlegm-damp, damp-heat, blood-stasis, qi-stagnation, and special constitutions. Some patients may have a combination of constitutions. In addition, detailed medical history data is obtained, including previous disease diagnosis records, treatment processes, medications, and family medical history. For TCM constitution characteristic data, feature extraction and quantitative analysis were performed based on the TCMQ questionnaire results; For medical history data, key information is extracted, including duration of illness, severity of disease, and treatment effect, and quantified, including classifying the severity of the disease into mild, moderate, and severe, and assigning corresponding numerical values ​​to each. These quantified TCM constitution characteristics and medical history data will be used together with the prediction results of the machine learning analysis module as the basis for subsequent evaluation; Construct an evaluation model. Based on a large amount of clinical experience data and medical research results, an expert-level evaluation algorithm model is constructed. The model uses a decision tree algorithm to take the patient's coronary heart disease risk prediction results, TCM constitution characteristic scores, and medical history data quantification values ​​as input features to construct a decision tree model. The nodes of the decision tree are divided according to the importance of different features. First, they are divided according to the coronary heart disease risk level, and then further subdivided according to the main TCM constitution types. Finally, the key factors in the medical history data, including whether there is a history of cardiovascular diseases such as hypertension and diabetes, are combined to determine the final evaluation result branch. Conduct model training and optimization, collect a large amount of sample data from patients with confirmed CHD and non-CHD, including their CHD risk prediction results, TCM constitution characteristics data and medical history data, as well as corresponding clinical diagnosis results; use these data to train the constructed evaluation model, and adjust the model parameters to make the model accuracy on the training set reach a high level; at the same time, use cross-validation method to evaluate the performance of the model on the validation set to avoid overfitting; if the model performance is poor, try to adjust the algorithm selection, increase the number of features, or use ensemble learning methods to optimize until the model meets the clinical application requirements in terms of accuracy, sensitivity and specificity; then provide targeted personalized health advice based on the recommended strategy based on the determined TCM constitution type of the patient.

[0006] Preferably, the adaptive filtering algorithm is implemented using a least mean square algorithm, which updates the filter coefficients by iteratively minimizing the error between the input signal and the expected signal, wherein the parameter setting includes setting the learning rate between 0.01 and 0.1 to achieve a balance between the filtering effect and the convergence speed; in addition, a recursive least squares algorithm is used to process signals in a high noise environment to further improve the signal enhancement effect.

[0007] Preferably, the process of updating the filter coefficients by iterative minimization is as follows: First, the initial coefficients of the filter are randomly set. Then, for each sampling moment, the input signal is filtered through the filter with the current coefficients to obtain the output signal. The error between the output signal and the expected signal is calculated, where the expected signal is the ideal ECG signal. The filter coefficients are updated using the formula based on the error and the input signal. At time n, the update formula of the filter coefficient w(n) is: w(n+1)=w(n)+μ*e(n)*x(n); where μ is the learning rate, which is set between 0.01 and 0.1; where e(n) is the error signal, and the error signal calculation formula is: e(n)=d(n)-y(n), d(n) is the expected signal, and y(n) is the filter output signal; x(n) is the input signal, that is, the actual collected ECG signal containing noise; by continuously iteratively updating the coefficients, the filter gradually adapts to the signal characteristics, suppresses noise, and enhances the useful information in the ECG signal; where the filter output formula is: ; Where L is the length of the filter, that is, the number of filter coefficients; is the i-th filter coefficient at time n; It's time The formula indicates that the output of the filter is the weighted sum of the input signal and the filter coefficient. By continuously adjusting the coefficient, the output signal is made closer to the expected signal, thereby enhancing the ECG signal. The adaptive filtering algorithm is implemented based on the least mean square algorithm. By continuously adjusting the filter coefficients, the error between the output of the input signal after filter processing and the expected signal is minimized. In ECG signal processing, the expected signal can be approximately considered as the original ECG signal without noise interference under ideal conditions, while the input signal is the actual collected ECG signal containing noise.

[0008] Preferably, based on a high noise environment, the filter coefficient update formula obtained by using a recursive least squares algorithm is: ; in, is the Kalman gain vector; is the inverse correlation matrix; is the forgetting factor, which ranges from 0.95 to 1 and is used to control the degree to which the algorithm forgets past data; It is a signal of expectation; is the input signal vector, which contains the input signal values ​​at the current time and several times in the past; It's time The filter coefficient vector of It's time Inverse correlation matrix; In a high noise environment, the noise interference of ECG signals is more serious, and the simple LMS algorithm may not achieve the ideal filtering effect; At this time, the recursive least squares algorithm is used for processing; The advantage of the RLS algorithm is that it can use all the input signal data in the past to calculate the filter coefficients at the current moment, instead of only considering the input and error at the current moment like the LMS algorithm; The recursive least squares algorithm RLS solves the filter coefficients that minimize the sum of weighted square errors by calculating the autocorrelation matrix of the input signal and the cross-correlation vector between the input signal and the expected signal; at each new sampling moment, it recursively updates the autocorrelation matrix and cross-correlation vector based on the new data, thereby quickly adapting to the dynamic changes of the signal; this algorithm can more accurately track signal changes in a high-noise environment, effectively remove noise, and improve the quality of ECG signals; including in the presence of strong electromagnetic interference or large patient motion artifacts, the RLS algorithm can better restore the true form of the ECG signal.

[0009] Preferably, the classification standards of TCM constitution characteristics include nine constitution types: balanced constitution, Qi deficiency constitution, Yang deficiency constitution, Yin deficiency constitution, phlegm-damp constitution, damp-heat constitution, blood stasis constitution, Qi stagnation constitution and special constitution. The expert-level evaluation algorithm performs feature extraction and analysis based on the TCM constitution classification questionnaire to ensure the comprehensiveness and accuracy of the evaluation.

[0010] Preferably, the voice prompts of the feedback module adopt WaveNet, a speech synthesis technology based on deep learning, to generate natural and fluent voice prompts; and the visual interface uses HTML5 and JavaScript technology to present the analysis results through a mobile application, including electrocardiogram curves, health scores and personalized health suggestions, so as to improve patients' understanding of their own health status.

[0011] Preferably, the wearable ECG acquisition device of the data acquisition module adopts high-sensitivity electrodes and a dynamic adjustment algorithm to ensure the quality of ECG acquisition in different body positions and motion states; the wearable ECG acquisition device is connected to the patient's mobile device via Bluetooth, and the patient views the collected ECG data in real time through a mobile application.

[0012] Preferably, the data preprocessing module includes two parts: signal filtering and baseline correction, wherein the signal filtering uses a multi-order low-pass filter to remove electromagnetic interference, and the baseline correction eliminates motion artifacts and baseline drift through wavelet transform to ensure the stability and accuracy of the ECG signal.

[0013] Preferably, the signal enhancement module adopts an adaptive filter to automatically adjust the filtering parameters according to the environmental noise characteristics detected in real time to minimize the impact of noise and improve the effectiveness of the ECG signal; the signal enhancement module also uses a deep learning algorithm to score the quality of the ECG signal. If the signal quality is found to be lower than the set score threshold, the patient is prompted to adjust the electrode position or change the collection posture through voice and text.

[0014] Preferably, the process of automatically adjusting the filtering parameters of the adaptive filter is as follows: First, real-time environmental noise detection and parameter adjustment are carried out, and the environmental noise characteristics are detected in real time through the adaptive filter in the signal enhancement module; the signal is converted to the frequency domain using the fast Fourier transform signal processing technology, and the distribution of noise in the frequency domain is analyzed to obtain the noise characteristics; the filter parameters are adjusted according to the frequency and amplitude in the noise characteristics; if strong interference noise of a specific frequency is detected in the environment, such as 50Hz or 60Hz power frequency interference, the filter can adjust its frequency response characteristics in a targeted manner to enhance the suppression ability of the frequency noise; at the same time, the gain parameters of the filter are adjusted according to the change of the noise amplitude to ensure effective filtering under different noise levels; including when the noise amplitude is detected to increase, the attenuation coefficient of the filter is appropriately increased to enhance the filtering effect of the noise; Then, the quality of ECG signals is evaluated using a deep learning algorithm. ECG signal samples of different quality levels are collected, including high-quality ECG signals and ECG signals under various noise interferences. ECG signal samples are labeled with high quality, medium quality, and low quality. A deep learning model is built based on a convolutional neural network. The collected ECG signal is input into the trained deep learning model. The model evaluates the signal quality based on the learned signal feature pattern and outputs a quality score. If the quality score is lower than the set score threshold, a voice announcement and text pop-up box will be generated and sent to the patient's mobile application, prompting the patient to adjust the position of the electrodes or change the collection posture through voice and text to re-collect high-quality ECG signals; then the quality of the newly collected signals will be evaluated again until the quality score of the signal quality is greater than or equal to the score threshold, which can ensure the accuracy of subsequent analysis and diagnosis; the voice announcement is: Please check whether the electrodes are firmly attached, and adjust the position of the electrodes if necessary; the content of the text pop-up box is the same as the voice announcement.

[0015] The data transmission module supports wireless communication protocols, including Bluetooth, Wi-Fi and cellular mobile networks, to ensure that data can be smoothly transmitted to the remote server in various environments; the AES-256 encryption algorithm is used to protect the security of the data during data transmission, and the data is transmitted in combination with the TLS transport layer security protocol to ensure the privacy and integrity of the data during transmission and prevent the leakage of patient privacy data.

[0016] The machine learning analysis module is constructed based on a convolutional neural network, wherein the model structure includes four convolutional layers and two pooling layers for extracting features of ECG signals; the convolutional layer uses a 3x3 convolution kernel, the activation function is ReLU, and the pooling layer uses a 2x2 maximum pooling; the training data comes from the public MIT-BIH arrhythmia database and annotated real patient data sets to ensure the effectiveness and accuracy of the model; the machine learning analysis module also uses transfer learning technology to initialize the current model by using a pre-trained ECG classification model to improve training efficiency and the generalization ability of the model; wherein the specific application scenario of transfer learning is to classify the ECG data of new patients by using a pre-trained ECG classification model; The ECG classification model is derived from a large-scale ECG dataset in the MIT-BIH arrhythmia database. The use of the pre-trained model can reduce the training time for new datasets and improve the prediction accuracy in different patient groups.

[0017] In addition to using the output of the machine learning model, the expert evaluation module also combines the TCM constitution classification standards and the patient's TCM prescription records to provide a more comprehensive and personalized assessment of coronary heart disease risk; the expert evaluation module is used to provide personalized health advice based on patients with different constitution types, including diagnostic advice and prevention plans corresponding to the patient's constitution type.

[0018] Specifically, for patients with a balanced constitution, it is recommended to maintain a balanced diet and exercise appropriately to maintain a healthy state; for patients with a qi deficiency constitution, it is recommended to consume more protein-rich foods and perform moderate qigong or tai chi exercises; for patients with a yang deficiency constitution, it is recommended to eat more warming foods, including ginger and mutton, and avoid cold environments; for patients with a yin deficiency constitution, it is recommended to drink more soups that nourish yin and moisten dryness, including tremella soup, and avoid spicy foods; for patients with phlegm-damp constitution, it is recommended to reduce the intake of greasy foods and perform aerobic exercise to promote metabolism; for patients with damp-heat constitution, it is recommended to eat more foods that clear heat and remove dampness, including coix seed and mung bean, and keep the environment ventilated; for patients with blood stasis constitution, it is recommended to perform meridian massage and Chinese medicine conditioning that activates blood circulation and removes blood stasis, including salvia miltiorrhiza and angelica sinensis; for patients with qi stagnation constitution, it is recommended to engage in relaxing activities, including yoga and meditation, and take Chinese medicine that soothes the liver and regulates qi appropriately; for patients with a special constitution, it is recommended to perform personalized conditioning according to specific allergies or special physical conditions, avoid allergens, and enhance immunity.

[0019] The feedback module includes a voice prompt system and a graphic display interface. The voice prompt system conveys the current heart health status and operation instructions to the patient in natural language; the graphic display interface displays the ECG analysis results, health advice and TCM diagnosis and treatment cooperation plan through the patient's mobile application.

[0020] The system includes a set of cloud computing-based back-end servers, which use distributed storage technology to store patients' historical data and analysis results; by using distributed storage technology, the efficiency of data processing and the reliability of the system are improved.

[0021] The machine learning analysis module analyzes the ECG data based on the convolutional neural network to predict the risk of coronary heart disease in patients. The process of obtaining the prediction result is as follows: ECG signal data with different characteristics, including normal ECG signals and ECG signals of various arrhythmia types, were obtained from the MIT-BIH arrhythmia database for model training. At the same time, ECG data of real patients in the hospital were collected and annotated by professional doctors. The annotated information included whether the patient had coronary heart disease and the specific type and severity of coronary heart disease, and the two parts of data were integrated to ensure the integrity and representativeness of the data. During the integration process, the data was formatted and cleaned to remove duplicate, erroneous or incomplete data records to ensure data quality, laying a solid foundation for subsequent model training and analysis. Preprocess the integrated ECG data, including signal filtering and normalization; when normalizing the signal, map the amplitude of the ECG signal to the interval [0,1] to eliminate the impact of different equipment acquisition or individual differences, and improve the stability and accuracy of model training; perform feature engineering to extract features related to the prediction of coronary heart disease risk; extract frequency domain features and time-frequency domain features through wavelet transform in signal processing technology; including using wavelet transform to decompose the ECG signal into sub-signals of different frequency bands, calculate the energy distribution of the sub-signals, and enrich the feature vector, so that the model can learn the relationship between the ECG signal and the risk of coronary heart disease more comprehensively; Construct a convolutional neural network architecture consisting of four convolutional layers and two pooling layers; the convolutional layer uses a 3x3 convolution kernel, which can effectively extract local features. As the convolutional layer goes deeper, it gradually learns more abstract and advanced feature representations; the activation function uses ReLU, which can accelerate the convergence speed of the model and effectively alleviate the gradient vanishing problem, so that the model can learn effective features faster during training; the pooling layer uses a 2x2 maximum pooling operation to downsample the feature map output by the convolutional layer, reduce the data dimension, reduce the amount of calculation, while retaining the main feature information, and improve the model's anti-interference ability and generalization ability; after the convolutional layer and the pooling layer, connect the fully connected layer for classification prediction; the fully connected layer maps the feature vector after convolution and pooling processing to the output layer. The number of nodes in the output layer is determined according to the prediction task and is used to predict the patient's coronary heart disease risk level, including low risk, medium risk, high risk, or whether the patient has coronary heart disease; Initialize the model using transfer learning technology; load some weight parameters from the pre-trained ECG classification model. These pre-trained models have learned some common ECG signal feature patterns, which can provide a good initial state for the current model, reduce the training time for new data sets, and improve the prediction accuracy in different patient groups; set the learning rate parameters of the model, the initial learning rate is set to 0.001, and dynamically adjust according to the learning curve during the training process, adopt the learning rate decay strategy, gradually reduce the learning rate as the number of training rounds increases to prevent the model from overfitting, set the model's regularization parameters including the L2 regularization coefficient, which is used to control the complexity of the model and prevent overfitting, and is set to 0.01; at the same time, select the Adam optimization algorithm from the appropriate optimization algorithm, which combines the advantages of the Adagrad and RMSProp algorithms, can adaptively adjust the learning rate, use different learning rates on different parameters, and accelerate the model convergence to train the model; The preprocessed ECG data is divided into training set, validation set and test set, with the ratio set to 70%:15%:15%; during the training process, the training set data is input into the convolutional neural network model in batches, and the model performs forward propagation to calculate the prediction results based on the input data, and then calculates the cross entropy loss function between the prediction results and the true labels; The gradient of the loss function to the model parameters is calculated through the back-propagation algorithm. The model parameters are updated according to the gradient and the selected optimization algorithm. The convolution kernel weights and bias parameters are continuously adjusted to make the model's prediction results gradually approach the true label. After each training round, the performance of the model is evaluated using the validation set data. The evaluation indicators of accuracy, recall rate, and F1 value are calculated to observe whether the model is overfitting or underfitting. If the model performance is poor on the validation set, including low accuracy or overfitting, a variety of tuning measures can be taken; adjust the model architecture, including increasing or decreasing the number of convolutional layers and pooling layers, and changing the size of the convolution kernel; increase the amount of data, and use data enhancement technology to generate more training samples by translating, rotating, and scaling the ECG signals to enrich data diversity; adjust hyperparameters: optimize the algorithm parameters and regularization coefficients; After multiple adjustments and training, the model's performance on the validation set reached a satisfactory level and also showed good generalization ability on the test set. At this point, the model has learned the effective features and patterns between ECG data and coronary heart disease risk, and can accurately predict the risk of new ECG data. When new patient ECG data is collected by the data acquisition module, processed by the preprocessing module, and enhanced by the signal enhancement module, it is input into the trained convolutional neural network model; the model performs forward propagation calculation on the input data, and outputs the prediction result according to the learned feature pattern and weight parameters; the prediction result is the probability value of the patient suffering from coronary heart disease, and the probability value is between 0 and 1, where the closer to 1, the higher the risk of disease, and it can also be a direct risk level classification, including low risk, medium risk, and high risk; The prediction results are passed to the expert evaluation module and retained in the machine learning analysis module as the basis for subsequent analysis and optimization of the model. In the expert evaluation module, the prediction results are combined with the patient's TCM constitution characteristics and medical history data for a more comprehensive and in-depth evaluation, ultimately generating personalized health advice and treatment plans. At the same time, the prediction results can also be displayed in a visual interface through the feedback module, with information such as risk level, probability value, and voice prompts in the patient's mobile application: feedback is provided to patients and doctors in the form of informing them of their risk status, so that they can promptly understand the patient's coronary heart disease risk status and take appropriate measures for prevention and treatment.

[0022] The present invention provides a coronary heart disease risk prediction system based on remote traditional Chinese medicine diagnosis and treatment, which has the following beneficial effects: This coronary heart disease risk prediction system based on remote traditional Chinese medicine diagnosis and treatment can, on the one hand, ensure the quality of ECG acquisition in different body positions and motion states based on high-sensitivity electrodes and dynamic adjustment algorithms; on the other hand, it can enhance the acquired signals through the signal enhancement module to restore the true characteristics of the ECG signals, so that subsequent analysis can be based on more accurate signals, helping machine learning models to more accurately extract features from ECG signals, improve measurement accuracy, and provide support for comprehensive assessments and personalized health recommendations. The entire system forms an organic whole, achieves collaborative work, and realizes effective prediction and management of coronary heart disease risks; it can also prompt patients to adjust the electrode position or change the acquisition posture through voice and text to further standardize patient operations and avoid the problem that the acquired ECG data contains a lot of noise and artifacts, which affects the accuracy of coronary heart disease risk prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of a coronary heart disease risk prediction system based on remote traditional Chinese medicine diagnosis and treatment according to the present invention. DETAILED DESCRIPTION

[0024] The present invention is further described in detail below in conjunction with the accompanying drawings and specific embodiments. The embodiments of the present invention are provided for the purpose of illustration and description, and are not intended to be exhaustive or to limit the present invention to the disclosed forms. Many modifications and variations will be apparent to those of ordinary skill in the art. The embodiments are selected and described in order to better illustrate the principles and practical applications of the present invention, and to enable those of ordinary skill in the art to understand the present invention and thereby design various embodiments with various modifications suitable for specific uses.

[0025] like Figure 1 As shown, the present invention provides a technical solution: a coronary heart disease risk prediction system based on remote traditional Chinese medicine diagnosis and treatment, comprising the following modules: a data acquisition module, a data preprocessing module, a signal enhancement module, a data transmission module, a machine learning analysis module, an expert evaluation module and a feedback module; The data acquisition module is used to obtain the patient's ECG signal through a wearable ECG acquisition device, and convert the analog signal into a digital signal through an analog-to-digital converter to obtain the ECG data. The ECG acquisition device is equipped with a remote communication function to achieve real-time data transmission and ensure the stability and accuracy of the acquisition process; The data preprocessing module is used to filter and denoise the collected ECG data to eliminate interference signals and ensure data quality; the signal enhancement module is used to enhance the preprocessed ECG signals through an adaptive filtering algorithm to improve signal clarity and stability; the data transmission module is used to upload the enhanced ECG data to the server via a wireless network, supporting a variety of wireless communication protocols, including Bluetooth, Wi-Fi, and cellular mobile networks, to ensure smooth data transmission; The machine learning analysis module analyzes the enhanced ECG data based on the convolutional neural network to predict the patient's risk of coronary heart disease and obtain the prediction results; The expert evaluation module is used to conduct expert-level evaluation based on the prediction results of the machine learning analysis module and the patient's TCM constitution characteristics and medical history data to obtain evaluation results and provide personalized health advice based on nine constitution types; The feedback module provides feedback of assessment results to patients and doctors through voice prompts and visual interfaces, and provides corresponding health advice.

[0026] The adaptive filtering algorithm is implemented using the least mean square algorithm, which updates the filter coefficients by iteratively minimizing the error between the input signal and the expected signal. The parameter setting includes setting the learning rate to 0.505 to strike a balance between filtering effect and convergence speed. In addition, the recursive least squares algorithm is used to process signals in high noise environments to further improve the signal enhancement effect.

[0027] The process of updating the filter coefficients by iterative minimization is as follows: First, the initial coefficients of the filter are randomly set. Then, for each sampling moment, the input signal is filtered through the filter with the current coefficients to obtain the output signal. The error between the output signal and the expected signal is calculated, where the expected signal is the ideal ECG signal. The filter coefficients are updated using the formula based on the error and the input signal. At time n, the update formula of the filter coefficient w(n) is: w(n+1)=w(n)+μ*e(n)*x(n); where μ is the learning rate, set to 0.505; where e(n) is the error signal, and the error signal calculation formula is: e(n)=d(n)-y(n), d(n) is the expected signal, and y(n) is the filter output signal; x(n) is the input signal, that is, the actual collected ECG signal containing noise; by continuously iteratively updating the coefficients, the filter gradually adapts to the signal characteristics, suppresses noise, and enhances the useful information in the ECG signal; where the filter output formula is: ; Where L is the length of the filter, that is, the number of filter coefficients; is the i-th filter coefficient at time n; It's time The formula indicates that the output of the filter is the weighted sum of the input signal and the filter coefficient. By continuously adjusting the coefficient, the output signal is made closer to the expected signal, thereby enhancing the ECG signal. The adaptive filtering algorithm is implemented based on the least mean square algorithm. By continuously adjusting the filter coefficients, the error between the output of the input signal after filter processing and the expected signal is minimized. In ECG signal processing, the expected signal can be approximately considered as the original ECG signal without noise interference under ideal conditions, while the input signal is the actual collected ECG signal containing noise.

[0028] Based on the high noise environment, the filter coefficient update formula obtained by recursive least squares algorithm is: ; in, is the Kalman gain vector; is the inverse correlation matrix; is the forgetting factor, which is set to 0.975 and is used to control the degree to which the algorithm forgets past data; It is a signal of expectation; is the input signal vector, which contains the input signal values ​​at the current time and several times in the past; It's time The filter coefficient vector of It's time Inverse correlation matrix; In a high noise environment, the noise interference of ECG signals is more serious, and the simple LMS algorithm may not achieve the ideal filtering effect; At this time, the recursive least squares algorithm is used for processing; The advantage of the RLS algorithm is that it can use all the input signal data in the past to calculate the filter coefficients at the current moment, instead of only considering the input and error at the current moment like the LMS algorithm; The recursive least squares algorithm RLS solves the filter coefficients that minimize the sum of weighted square errors by calculating the autocorrelation matrix of the input signal and the cross-correlation vector between the input signal and the expected signal; at each new sampling moment, it recursively updates the autocorrelation matrix and cross-correlation vector based on the new data, thereby quickly adapting to the dynamic changes of the signal; this algorithm can more accurately track signal changes in a high-noise environment, effectively remove noise, and improve the quality of ECG signals; including in the presence of strong electromagnetic interference or large patient motion artifacts, the RLS algorithm can better restore the true form of the ECG signal.

[0029] The classification standards for TCM constitution characteristics include nine constitution types: balanced constitution, qi deficiency constitution, yang deficiency constitution, yin deficiency constitution, phlegm-dampness constitution, damp-heat constitution, blood stasis constitution, qi stagnation constitution and special constitution. The expert-level assessment algorithm extracts and analyzes features based on the TCM constitution classification questionnaire to ensure the comprehensiveness and accuracy of the assessment.

[0030] The voice prompts of the feedback module use WaveNet, a deep learning-based speech synthesis technology, to generate natural and fluent voice prompts; while the visual interface uses HTML5 and JavaScript technology to present analysis results through mobile applications, including ECG curves, health scores and personalized health recommendations, to improve patients' understanding of their own health status.

[0031] The generated expert-level evaluation results are passed to the feedback module; the feedback module presents this information to patients and doctors through a voice prompt system and a graphic display interface; the voice prompt system conveys the evaluation results to patients in a natural and fluent voice, including informing patients that "according to the comprehensive evaluation, your current risk of coronary heart disease is at a medium risk level, and your constitution type is phlegm-dampness. It is recommended that you reduce your intake of greasy food and increase exercise", so that patients can intuitively understand their physical condition; the graphic display interface displays the evaluation results in detail in the patient's mobile application, including a chart showing the trend of changes in the probability of coronary heart disease risk, the score distribution of TCM constitution types, and a detailed text listing of the specific content of personalized health recommendations, so that patients can view and refer to them at any time; at the same time, doctors can also obtain this information through the feedback module for reference in subsequent diagnosis and treatment, providing patients with more comprehensive medical services.

[0032] The wearable ECG acquisition device of the data acquisition module adopts high-sensitivity electrodes and dynamic adjustment algorithms to ensure the quality of ECG acquisition in different body positions and motion states; the wearable ECG acquisition device is connected to the patient's mobile device via Bluetooth, and the patient can view the collected ECG data in real time through the mobile application.

[0033] The data preprocessing module includes two parts: signal filtering and baseline correction. Signal filtering uses a multi-order low-pass filter to remove electromagnetic interference, and baseline correction eliminates motion artifacts and baseline drift through wavelet transform to ensure the stability and accuracy of the ECG signal.

[0034] The signal enhancement module uses an adaptive filter to automatically adjust the filtering parameters according to the real-time detected environmental noise characteristics to minimize the impact of noise and improve the effectiveness of the ECG signal. The signal enhancement module also uses a deep learning algorithm to score the quality of the ECG signal. If the signal quality is found to be lower than the set score threshold, the patient is prompted through voice and text to adjust the electrode position or change the collection posture.

[0035] The process of automatically adjusting the filtering parameters of the adaptive filter is as follows: First, real-time environmental noise detection and parameter adjustment are carried out. The environmental noise characteristics are detected in real time through the adaptive filter in the signal enhancement module. The signal is converted to the frequency domain using the fast Fourier transform signal processing technology, and the distribution of noise in the frequency domain is analyzed to obtain the noise characteristics. The filter parameters are adjusted according to the frequency and amplitude in the noise characteristics. If strong interference noise of 50Hz is detected in the environment, the filter can adjust its frequency response characteristics in a targeted manner to enhance the suppression ability of the noise of this frequency. At the same time, the gain parameters of the filter are adjusted according to the change of the noise amplitude to ensure effective filtering under different noise levels. This includes appropriately increasing the attenuation coefficient of the filter when the noise amplitude is detected to enhance the filtering effect of the noise. Then, the quality of ECG signals is evaluated using a deep learning algorithm. ECG signal samples of different quality levels are collected, including high-quality ECG signals and ECG signals under various noise interferences. ECG signal samples are labeled with high quality, medium quality, and low quality. A deep learning model is built based on a convolutional neural network. The collected ECG signals are input into the trained deep learning model, and the model evaluates the signal quality based on the learned signal feature pattern and outputs a quality score. If the quality score is lower than the set score threshold, a voice announcement and text pop-up box will be generated and sent to the patient's mobile application, prompting the patient to adjust the position of the electrodes or change the collection posture through voice and text to re-collect high-quality ECG signals; then the quality of the newly collected signals will be evaluated again until the quality score of the signal quality is greater than or equal to the score threshold, which can ensure the accuracy of subsequent analysis and diagnosis; the voice announcement is: Please check whether the electrodes are firmly attached, and adjust the position of the electrodes if necessary; the content of the text pop-up box is the same as the voice announcement.

[0036] It should be further explained that, in the specific implementation process, the data acquisition module uses high-sensitivity electrodes and dynamic adjustment algorithms to ensure the quality of ECG acquisition in different body positions and motion states; however, even in the case of high-quality acquisition, a certain degree of noise is still inevitable; the signal enhancement module works closely with the data acquisition module to enhance the signal immediately after it is acquired; the adaptive filtering algorithm is used to quickly remove these noises and restore the true characteristics of the ECG signal, so that subsequent analysis can be based on more accurate signals.

[0037] The high-quality ECG signal processed by the signal enhancement module is transmitted to the machine learning analysis module; in the machine learning analysis module, the ECG data is analyzed based on the convolutional neural network to predict the patient's risk of coronary heart disease; because the signal enhancement module effectively removes noise interference, the machine learning model can more accurately extract the features in the ECG signal; these accurate features are crucial for accurately predicting the risk of coronary heart disease; and the clear signal after signal enhancement can improve the accuracy of measurement, thereby improving the accuracy of coronary heart disease risk prediction; at the same time, the prediction results of the machine learning analysis module provide an important basis for the expert evaluation module, combined with the patient's TCM constitution characteristics and medical history data, to provide support for more comprehensive evaluation and personalized health advice. The entire system forms an organic whole, working together to achieve effective prediction and management of coronary heart disease risks.

[0038] Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field and related fields without creative work should fall within the scope of protection of the present invention. The structures, devices and operating methods not specifically described and explained in the present invention are implemented according to the conventional means in the field unless otherwise specified and limited.

Claims

1. A coronary heart disease risk prediction system based on remote traditional Chinese medicine diagnosis and treatment, characterized in that: It includes the following modules: data acquisition module, data preprocessing module, signal enhancement module, data transmission module, machine learning analysis module, expert evaluation module and feedback module; The data acquisition module is used to acquire the patient's ECG signal through a wearable ECG acquisition device, and convert the analog signal into a digital signal through an analog-to-digital converter to obtain ECG data. The ECG acquisition device is equipped with a remote communication function to achieve real-time data transmission; The data preprocessing module is used to filter and denoise the collected ECG data; the signal enhancement module is used to enhance the preprocessed ECG signal through an adaptive filtering algorithm; the data transmission module is used to upload the enhanced ECG data to the server through a wireless network; The machine learning analysis module analyzes the enhanced ECG data based on a convolutional neural network to predict the patient's risk of coronary heart disease and obtain a prediction result; The expert evaluation module is used to perform expert-level evaluation based on the prediction results of the machine learning analysis module and in combination with the patient's TCM constitution characteristics and medical history data to obtain evaluation results and provide personalized health advice based on the nine constitution types; The feedback module feeds back the evaluation results to the patient and the doctor through voice prompts and a visual interface, and provides corresponding health advice.

2. The coronary heart disease risk prediction system based on remote TCM diagnosis and treatment according to claim 1 is characterized by: The adaptive filtering algorithm is implemented using the least mean square algorithm, which updates the filter coefficients by iteratively minimizing the error between the input signal and the expected signal. The parameter setting includes setting the learning rate between 0.01 and 0.1 to achieve a balance between filtering effect and convergence speed. In addition, a recursive least squares algorithm is used to process signals in high noise environments.

3. The coronary heart disease risk prediction system based on remote TCM diagnosis and treatment according to claim 2 is characterized by: The process of updating the filter coefficients by iterative minimization is as follows: First, the initial coefficients of the filter are randomly set. Then, for each sampling moment, the input signal is filtered through the filter with the current coefficients to obtain the output signal. The error between the output signal and the expected signal is calculated, where the expected signal is the ideal ECG signal. The filter coefficients are updated using the formula based on the error and the input signal. At time n, the update formula of the filter coefficient w(n) is: w(n+1)=w(n)+μ*e(n)*x(n); where μ is the learning rate, set between 0.01 and 0.1; where e(n) is the error signal, and the error signal calculation formula is: e(n)=d(n)-y(n), d(n) is the expected signal, y(n) is the filter output signal; x(n) is the input signal, that is, the actual collected ECG signal containing noise; where the filter output formula is: ; Where L is the length of the filter, that is, the number of filter coefficients; is the i-th filter coefficient at time n; It's time The input signal.

4. The coronary heart disease risk prediction system based on remote TCM diagnosis and treatment according to claim 3 is characterized by: Based on the high noise environment, the filter coefficient update formula obtained by recursive least squares algorithm is: ; in, is the Kalman gain vector; is the inverse correlation matrix; is the forgetting factor, which ranges from 0.95 to 1 and is used to control the degree to which the algorithm forgets past data; It is a signal of expectation; is the input signal vector, which contains the input signal values ​​at the current time and several times in the past; It's time The filter coefficient vector of It's time The inverse correlation matrix of .

5. A coronary heart disease risk prediction system based on remote TCM diagnosis and treatment according to claim 4, characterized in that: The classification standards for TCM constitution characteristics include nine constitution types: balanced constitution, qi deficiency constitution, yang deficiency constitution, yin deficiency constitution, phlegm-dampness constitution, damp-heat constitution, blood stasis constitution, qi stagnation constitution and special constitution. The expert-level evaluation algorithm performs feature extraction and analysis based on the TCM constitution classification questionnaire.

6. A coronary heart disease risk prediction system based on remote TCM diagnosis and treatment according to claim 5, characterized in that: The voice prompts of the feedback module adopt WaveNet, a speech synthesis technology based on deep learning, to generate natural and fluent voice prompts; The visualization interface uses HTML5 and JavaScript technology to present analysis results, including ECG curves, health scores, and personalized health recommendations, through a mobile application.

7. A coronary heart disease risk prediction system based on remote TCM diagnosis and treatment according to claim 6, characterized in that: The wearable ECG acquisition device of the data acquisition module adopts high-sensitivity electrodes and a dynamic adjustment algorithm to ensure the quality of ECG acquisition in different body positions and motion states; the wearable ECG acquisition device is connected to the patient's mobile device via Bluetooth, and the patient can view the collected ECG data in real time through a mobile application.

8. The coronary heart disease risk prediction system based on remote TCM diagnosis and treatment according to claim 7 is characterized by: The data preprocessing module includes two parts: signal filtering and baseline correction. The signal filtering uses a multi-order low-pass filter to remove electromagnetic interference, and the baseline correction eliminates motion artifacts and baseline drift through wavelet transformation.

9. The coronary heart disease risk prediction system based on remote TCM diagnosis and treatment according to claim 8, characterized in that: The signal enhancement module adopts an adaptive filter to automatically adjust the filtering parameters according to the environmental noise characteristics detected in real time; the signal enhancement module also uses a deep learning algorithm to score the quality of the ECG signal. If the signal quality is found to be lower than the set score threshold, the patient is prompted to adjust the electrode position or change the collection posture through voice and text.

10. A coronary heart disease risk prediction system based on remote TCM diagnosis and treatment according to claim 9, characterized in that: The process of automatically adjusting the filtering parameters of the adaptive filter is as follows: First, real-time environmental noise detection and parameter adjustment are performed. The environmental noise characteristics are detected in real time through the adaptive filter in the signal enhancement module. The signal is converted to the frequency domain using the fast Fourier transform signal processing technology, and the distribution of noise in the frequency domain is analyzed to obtain the noise characteristics. The filter parameters are adjusted according to the frequency and amplitude of the noise characteristics. Then, the quality of ECG signals is evaluated using a deep learning algorithm. ECG signal samples of different quality levels are collected, including high-quality ECG signals and ECG signals under noise interference. ECG signal samples are labeled with high quality, medium quality, and low quality. A deep learning model is built based on a convolutional neural network. The collected ECG signal is input into the trained deep learning model. The model evaluates the signal quality based on the learned signal feature pattern and outputs a quality score. If the quality score is lower than the set score threshold, a voice announcement and text pop-up box will be generated and sent to the patient's mobile application, prompting the patient through voice and text to adjust the electrode position or change the collection posture to re-collect high-quality ECG signals; then the quality of the newly collected signal will be evaluated again until the quality score of the signal quality is greater than or equal to the score threshold.

Images