Multiscale energy based suitable wavelet selection for detection of myocardial infarction in ECG

Abstract

Over the decades, electrocardiogram (ECG) has been proved as the chief diagnostic tool for assessment of the cardiovascular condition of human being. Myocardial Infarction (MI) is commonly known as heart attack, happens when blood supply stops to heart muscles causing occlusions in some portion or whole artery. MI is the result of three pathological changes such as elevation of ST-segment, the appearance of wide pathological Q-wave and inversion of T-wave in ECG record. Detection of MI by considering few ECG leads generally requires prior information about the pathological behaviour of the disease. The present work considers 12 leads to view the cardiac condition from various angles in ECG signal for accurate detection of MI. This Letter investigates on various wavelet basis functions, i.e. Haar, Daubechies, Symlet, Coiflet and biorthogonal basis filters of different order for selecting the most suitable one for the detection of MI. Wavelet transform of 12-lead ECG signal decomposes the signal into different subbands. A comparative study has been done based on the multiscale energy at different wavelet subbands for the selection of most suitable wavelet basis for the accurate detection of MI. The experimentation is carried out on different datasets from the PTB diagnostic ECG database.

Keywords

• blood vessels
• cardiovascular system
• cardiology
• diseases
• muscle
• filtering theory
• medical signal processing
• wavelet transforms
• electrocardiography
• multiscale energy
• suitable wavelet selection
• myocardial infarction
• chief diagnostic tool
• cardiovascular condition
• MI
• heart attack
• heart muscles
• pathological changes
• wide pathological Q-wave
• ECG record
• pathological behaviour
• cardiac condition
• accurate detection
• wavelet basis functions
• wavelet transform
• 12-lead ECG signal
• different wavelet subbands
• suitable wavelet basis
• PTB diagnostic ECG database