Characterization of post MI electrocardiogram using power ratio features and K-nearest neighbor classifier

Abstract

Myocardial infarction (MI) is the irreversible necrosis of heart muscles caused by prolonged ischemic condition. Subsequently the presence of damaged tissues in post-MI patients is expected to have an effect on their electrocardiogram (ECG). Hence, this paper proposes characterization of post-MI ECG from bipolar and augmented limb leads. Data from patients of inferior and anterior MI, as well as healthy controls are acquired from PTB Diagnostic ECG Database. Noise removal is performed using band-pass filter. Power ratio is then extracted from energy spectral density of bipolar and augmented limb leads. Classification via k-nearest neighbor method revealed that features from the augmented limb leads outperform bipolar limb leads in terms of training and testing accuracies. Hence, power ratio features from the augmented limb leads exhibit higher degree of separation between the ECG classes.

Keywords: myocardial infarction; electrocardiogram; power ratio; k-nearest neighbor