Biomedical Signal Processing For Medical Diagnosis

Abstract

A classification system utilizing the nearest neighbor (NN) method has been created to screen for Antero Septal Myocardial Infarction (ASMI). In this approach, we extract QRS amplitude and T height features—both of which are crucial for diagnosis—from leads V1 to V4. To combine the effects of these four leads, we employ a scoring method. Interestingly, both Euclidean and Mahalanobis distance metrics are applied in the NN classifier, with the latter showing better performance. Additionally, time-plane ECG feature-based classification methods rely on explicit time-plane features, resulting in a large set of features. By applying a carefully crafted mathematical formula, we can extract key parameters from the wavelet cross spectrum and wavelet coherence to identify both normal and abnormal cardiac patterns, like Inferior MI. To ensure our approach is reliable, we conducted empirical tests that yielded an impressive accuracy rate of 97.6% when tested on our database.