The Analysis And Processing Of ECG Signal And The Establishment Of Myocardial Infarction Model

Myocardial infarction is one of the most common heart diseases and electrocardiogram(ECG) examination is the primary method used to diagnose it. Because of the individuality and diversity of its EGC features for myocardial infarction, it is difficult to achieve the ideal diagnosis effect. It’s significant for patients to identify the specific features of myocardial infarction and take steps timely. ECG technology has got increasingly mature after more than one hundred years’ development and ECG machine has got more precision and perfect as the development of electronics technology. However, the diagnosis effect of the clinical doctor useing ECG to diagnose diseases was far from ideal. Currently, most of diagnostic criteria come from the diagnostic experience of doctors. It is common that there is a lot of controversy on the application for some diseases of some indexes. Therefore, it has become the key points of today’s research to dig out the specific features which could be applied to the cardiovascular disease diagnosis, to overcome the shortcomings of ECG, and to increase the accuracy and efficiency of ECG diagnosis, through analyzing and processing the ECG information obtained from computer technology,The paper started the research from the ECG signals of myocardial infarction patients and healthy control. First, the signals of 158 myocardial infarction patients and 90 healthy volunteers were collected from the Physiobank, then the data were pre-processed by means of median filtering, band-stop filtering and wavelet decomposition, in whcih the baseline shift noise, power frequency disturbance and high-frequency noise were removed. Second, 11 features of ECG were extracted by means of wavelet transform, differential threshold method, slope-threshold arithmetic and so on and the 11 features of ECG were the amplitude of R-waves, the time width of QRS complex wave, RR interval, the amplitude and time width of P-waves, the amplitude and time width of T-waves, PR interval, QT interval, RR interval, ST potential migration and premature beat. Finally, eight parameters which were much correlated with diagnosis were selected by means of Pearson correlation index for the establishment of assistant diagnose model of myocardial infarction, and the performance of the models was evaluated and the physiological significance of the parameters was elaborated as well.The main contribution and innovation of this article are as follows:(1).The methods of pro-processing of ECG signals were systematically studied. The best width of median filter window and parameters like the wavelet basis and decomposition layer-number of wavelet filter removing the high-frequency noise in ECG were obtained through experimental simulation and analysis.(2). The technology of ECG feature extracting was analyzed and realized, 11 features including ST potential migration of ECG were extracted.(3). The linear and nonlinear myocardial infarction models were established. It was concluded that the accuracy of SVM classifier was higher than linear regression model, and premature beat, weakness amplitude of R-waves, T wave inversion, longer QT interval, ST potential migration were specific symbols and risk factors of myocardial infarction, which provided reference for the clinical diagnosis of MI.