Research And Application Of Auxiliary Diagnosis Algorithm For Myocardial Infarction Based On Machine Learning

Myocardial infarction is a serious class of cardiovascular diseases,and its onset is characterized by easy sudden onset,dangerous condition,easy to cause death and easy to cause related diseases,etc.Pre-diagnosis and early diagnosis of myocardial infarction are of great significance to the prevention,treatment and prognosis of patients.The current common studies using ECG automatic diagnosis technology for myocardial infarction assisted diagnosis often have problems such as many feature points to be localized,low accuracy of feature point localization except for QRS wave groups,feature extraction heavily relying on the accuracy of feature point localization,feature extraction easily ignoring weak signal features,poor performance of classification diagnosis,and lack of out-of-hospital prediagnosis and assisted diagnosis systems.This paper focuses on the machine learning-based myocardial infarction assisted diagnosis algorithm and its application,and seeks to simplify the ECG signal processing and feature extraction process in the ECG-based infarction assisted diagnosis process and improve the classification effect of the infarction assisted diagnosis model.The main contents of the paper are as follows:(1)Research on ECG signal processing and feature point localization related methods.Based on the analysis of ECG signal interference,the filter group method is used to filter out the baseline drift,industrial frequency interference and EMG noise in the signal.In order to address the problems of many feature localization points and poor accuracy of some feature localization in the traditional ECG signal feature localization process,an adaptive thresholdbased ECG signal feature localization and feature fragment selection method is designed.The method uses adaptive threshold is used to locate the S-valley of ECG segments and crop the feature segments,which simplifies the process of feature point localization and lays the foundation for subsequent feature extraction.(2)To study the wavelet coefficient method based on multiscale analysis to extract the features of myocardial infarction ECG signals.Based on the multi-resolution analysis of infarct ECG signal features and wavelet transform,this paper proposes the method of multiscale analysis of feature fragments,directly using wavelet transform coefficients at multiresolution as feature vectors,using db8 wavelets for feature fragments with scale 4,and using decomposed wavelet coefficients as features.The method can fully and effectively retain and identify the weak features of the signal and does not have to rely on the localization accuracy of the many feature points of the signal.The experimental classification results show that the features extracted by this method have good characterization ability for myocardial infarction ECG signal features.(3)To study the machine learning-based assisted diagnosis model for myocardial infarction.Based on the ECG signal of PTB database,the Bagging-Trees integrated learning method is used as the main method,while decision tree,linear discriminant,logistic regression,support vector machine and K-nearest neighbor methods are used for the auxiliary diagnostic analysis of myocardial infarction and cardiovascular diseases.The experimental results show that the Bagging-Trees integrated learning method has better performance and better model classification for the auxiliary diagnosis of myocardial infarction compared with other machine learning methods;the Bagging-Trees integrated learning method also performs well for the auxiliary diagnosis of cardiovascular diseases such as myocardial infarction.(4)Design and implementation of myocardial infarction assisted diagnosis system.Based on the machine learning myocardial infarction assisted diagnosis model,this paper designs a front-and back-end separated assisted diagnosis system for out-of-hospital and other nonprofessional medical conditions,and uses hybrid programming to implement the assisted diagnosis server platform and user terminal with a displayable collaborative work interface.The experimental tests of the system show that the system is practical and can meet the needs of prediagnosis and assisted diagnosis of myocardial infarction under certain conditions.