Research On Feature Extraction And The Auxiliary Diagnosis Model Of Myocardial Infarction

With the change of people’s living standards and way of life,morbidity and mortality of cardiovascular disease increase year by year,accompanied by a trend of younger type.As one of the greatest causes of cardiovascular disease is a major disease that threatens human life.At present,the diagnosis of cardiac infarction is mainly focus on two method:the level of myocardial marker and ECG analysis.However,the extraction and analysis of the former are so complicated that it couldn’t be finished at short notice,so the automatic analysis of ECG has become the focus of experts and scholars research.This paper is based on myocardial infarction data,summarize previous research experience,research the pretreatment of the ECG noise、detection of ECG waveform and establishment of myocardial infarction diagnosis mode in detail.(1)The pretreatment of the ECG noise: a filter bank which can filter three kinds of interference is designed in this paper,according to the source of ECG noise and the requirement of real time analysis.This method not only can effectively filter the noise of different frequency such as the low frequency baseline drift、power frequency interference and EMG interference,but also does not affect the important information of the original signal.(2)Detection of ECG waveform: On the basis of multi-scale transform of continuous wavelet,an improved wavelet algorithm was proposed in this paper to detect ECG characteristic waveform by adaptively changing the threshold and increase the window function,which greatly improves the accuracy of feature extraction.The selection of wavelet basis function effectively suppresses the residual noise in ECG signal preprocessing to a great extent,the choice of wavelet scale and adaptive threshold is determined by a lot of experiments.The algorithm is verified by using some data of PTB diagnostic database,and the simulation results show that the accuracy rate of the R wave detection is more than 99%,which can accurately locate the ECG signal of serious interference and waveform distortion.(3)The establishment of myocardial infarction diagnosis mode: Based on the characteristic parameters of ECG signal such as amplitude of typical wave、interval and potential shift,Logistic regression、BP neural network and Support vector machine based on K-CV three kinds of automatic classification diagnosis model are designed.In the process of modeling,the optimum of c and g of SVM is studied,and a method of 20 percent off K-CV optimization for two times is proposed,which improves the performance of the classifier to a great extent.Finally,the performance of these three models is verified,the results show that the accuracy rate of the SVM model based on K-CV is 99.19%,which provide theoretical guidance and important clinical significance for diagnosis of myocardial infarction disease.