Research On Auxiliary Diagnosis Algorithm Of Myocardial Infarction

Myocardial infarction is myocardial necrosis caused by acute and persistent ischemia and hypoxia of coronary arteries,and has the characteristics of sudden and lethal.ECG signals of patients with myocardial infarction have specific changes,and the ECG signal detection has the characteristics of convenience and non-invasiveness.The auxiliary diagnosis algorithm of myocardial infarction based on ECG signal is of great significance for the early prevention and intervention of patients with myocardial infarction.This article focuses on the research on the myocardial infarction assisted diagnosis algorithm,mainly to solve the following shortcomings of the current myocardial infarction assisted diagnosis algorithm based on ECG signals: During the process of ECG signal denoising,the low frequency noise is close to the ST segment frequency band,affecting the ST segment morphology The detection accuracy of the QRS wave of the electrical signal is not high;The ECG feature of myocardial infarction diagnosis is single,and the diagnostic effect of myocardial infarction is poor due to differences in inter-patient.In view of the above shortcomings,this paper studies ECG signal preprocessing,feature extraction and myocardial infarction auxiliary diagnosis algorithm,the main research results are as follows :1.A low-frequency noise filtering method based on empirical mode decomposition is proposed to effectively filter the baseline drift caused by the low-frequency noise of the electrocardiogram signal and the interference of similar frequency components in the ST segment.The noise of the ST segment and the baseline drift has similar frequency components.Direct filtering of low-frequency signals will cause distortion of the ST segment.In this paper,empirical mode decomposition is used,the intrinsic mode function of frequency component of 0 to 1 Hz exceeding 50% are set to zero to achieve low-frequency noise filtering.Use stationary wavelet transform on the signal,and complete the filtering of power frequency interference and high frequency noise by zeroing the high frequency detail coefficients.2.An R wave peak point detection algorithm based on S transform and correlation entropy envelope is designed,which effectively reduces the false detection rate of QRS wave detection.For R-wave peak point detection,firstly,the filtered ECG signal is overlapped and overlapped to realize the S transform,then calculate the correlation entropy envelope to enhance the QRS wave information,and finally obtain the R wave peak point by adaptively adjusting the distance and amplitude of the peak point.109,642 heartbeats in the MIT-BIH arrhythmia database were used for the experiment.The results showed that the sensitivity,positive detection and accuracy of the R-wave peak point detection reached 99.86%,99.86% and 99.73% respectively,and the heart beat segmentation is realized based on the R wave peak point.3.A multi-feature fusion method of 12-lead ECG signal statistical features,entropy features and deep network features is proposed,which effectively improves the express ability of ECG signal features.Effectively extracting the key features that can characterize the ECG signal of myocardial infarction is the premise of auxiliary diagnosis.This article mainly extracts the two aspects' features of 12-lead ECG.The first type is to manually extract the features that can reflect the characteristics of the signal,including the mean,standard deviation,kurtosis coefficient,skewness coefficient and entropy features,reflecting the shape,amplitude,time and spectrum complexity and small abnormal dynamic changes of the QRS wave,ST segment and T wave in the heart beat.The other type is the deep network features,which uses the residual neural network to extract the hidden internal representation information of the heart beat of the ECG signal.On this basis,this paper proposes a multi-feature fusion method of artificial features and deep network features to extract a total of 352 features in 12 leads.4.Random forest,support vector machine and K-nearest neighbor algorithm are used to realize the auxiliary diagnosis of myocardial infarction in two modes of inter-patient and intra-patient.On the basis of multi-feature fusion of ECG signals,this paper based on the PTB myocardial infarction database,random forest,support vector machine and K-nearest neighbor algorithm were used to diagnose myocardial infarction,and the diagnosis results of intra-patient and inter-patient myocardial infarction were analyzed.The results show that: Compared with other classifiers,the random forest classifier has the best detection results in both inter-patient and intra-patient modes.The accuracy,sensitivity,specificity and F1 value of the diagnosis of myocardial infarction in the intra-patient are 99.95%;The accuracy,sensitivity,specificity and F1 value of the diagnosis of myocardial infarction in the inter-patient were 93.89%,99.54%,88.25%and 94.22%,respectively.Compared with the existing research results,the algorithm proposed in this paper has significantly improved the myocardial infarction diagnostic performance in intra-patient and inter-patient modes.