An Efficient Method for Detection and Localization of Myocardial Infarction

Abstract

Electrocardiography (ECG) is a noninvasive diagnostic tool used to diagnose coronary artery disease, myocardial infarction (MI), and other heart disorders. As MI is the death of cardiac muscle tissue owing to a coronary artery blockage. It should be diagnosed at an early stage to provide timely treatment and thereby save lives. This study proposes a novel efficient technique for MI detection and localization based on variational mode decomposition (VMD) and regularized neighborhood component analysis (RNCA). Statistical and nonlinear features calculated from intrinsic mode functions decomposed by VMD create the final feature set. These features are ranked using RNCA, a nonparametric feature ranking approach, and then fed into the k-nearest neighbors (KNN) and AdaBoost classifiers with minimum features. With 12-Lead ECG System, i.e., ten electrodes on the body for detection might be costly and restricts patient movement. The Physikalisch-Technische Bundesanstalt ECG diagnostic database consisting of all 12 leads data is studied to find the minimum number of leads needed for successful MI diagnosis and localizing infarcted artery. Using 33 features from lead 8 (V2) with the KNN, the proposed approach gave the best accuracy of 99.82% for detection compared to previous related studies. The technique also distinguished the 11 types of MI with 99.75% accuracy utilizing 22 features from lead 7 (V1) with the KNN. The study reveals that an automated diagnostic tool for the portable device may be built just by chest lead. As VMD provides a robust solution against noise, this algorithm is excellent for portable health devices.