Analysis of features for myocardial infarction and healthy patients based on wavelet

Abstract

An electrocardiogram (ECG) is the recording of the electrical activity of the heart. For different pathologies, different changes are observed in a normal ECG signal. In this paper, the features of 12 lead ECG signals are analyzed using wavelet decomposition and eigen space analysis for the detection. Wavelet decomposition distributes the diagnostic information present in the ECG signal amongst different sub-bands. It is observed that changes in the ECG signal for myocardial infarction patients are reflected in the statistical parameters (mean, variance, standard deviation and entropy) and wavelet energies of the wavelet coefficients of each sub-band that are calculated after decomposition with the mother wavelet and also in the eigen values calculated from covariance matrices obtained from subband matrices in the eigen space. Therefore, the statistical parameters along with wavelet energies and the eigen values can be used as training features for classification of the ECG signals into those belonging to that of healthy control (HC) and myocardial infarction (MI) patients. The 12 lead ECG signals of both healthy control (HC) and myocardial infarction (MI) are obtained from the PTB Diagnostic ECG Database.