Research On The Diagnosis Method Of Cardiovascular Disease Based On 12-lead ECG Signal

With the improvement of living standards and changes in lifestyle,cardiovascular disease has become a major disease threatening the health of our people,and the number of deaths due to cardiovascular disease is increasing every year.In the study of cardiovascular diseases,ECG data is a very important part.Through the abnormal detection of ECG data,early cardiovascular diseases can be identified and prevented and treated.Therefore,the detection and research of ECG data is extremely important.Due to the limited number of medical experts,it is impossible to process the massive amount of ECG data generated in a timely manner.Therefore,it is particularly important to study high-accuracy automatic analysis and diagnosis methods.The automatic analysis and diagnosis method of convolution based on ECG signals can be divided into four steps of signal acquisition,data preprocessing,feature extraction and classification recognition according to the process.This thesis takes 12-lead ECG signal as the research object,and mainly studies the feature extraction method and classification recognition method of ECG signal.The main research work can be summarized as the following two parts.The first part is the research of feature extraction method.Traditional feature extraction methods require manual segmentation of heart beats,which requires a large amount of engineering and is not ideal.However,the feature extraction method based on deep learning has the problem that it is difficult to take into account both the deep features and the time-series features of the ECG signal,which may result in poor detection of some diseases.Based on this,in order to extract more accurate comprehensive features of ECG data,this thesis discusses the construction of a single-lead ECG signal feature extraction model,which combines the local features of ECG signals with the timing features,so that the extracted The characteristics also include the complete characteristics of the ECG signal.The method has proved its effectiveness through experiments.The second part is the study of classification and recognition methods.The traditional classification and recognition methods mainly aim at single-lead ECG signals for classification and recognition,but the information provided by single-lead ECG signals is extremely limited and can no longer meet the needs of disease diagnosis,so disease diagnosis based on 12-lead ECG signals is required.The 12-lead ECG signal contains more information about the patient,but the feature information and feature positions involved are complex.How to effectively use the effective information in the 12-lead ECG signal has become a research difficulty.Based on this,this thesis builds a model based on a multi-layer attention mechanism,which takes into account the characteristics of multi-lead ECG signals,and at the same time pays attention to the characteristics of leads and frequencies of different diseases.The experimental results show that the method in this thesis aims at an average F1 score of 8diseases classified by 12-lead ECG signals to reach 0.882,and achieves high-accuracy diagnosis of diseases based on 12-lead ECG signals.