Multimodal Classification of Arrhythmia and Ischemia Using QRS-ST Analysis

Abstract

Probabilistic classification approaches have been presented for arrhythmic and ischemic data using QRS-ST evaluation. The proposed methodology is segregated into two major parts, i.e., (a) detection of QRS complex and ST segments by improvised Pan-Tompkins and difference operation method, respectively, and (b) classification of healthy, arrhythmic, and ischemic classes using linear discriminant analysis (LDA), decision tree (DT), and artificial neural network (ANN), respectively. Two correlative classification features (frequency and time domain) of QRS-ST, i.e., (1) ratio of power spectrum (PS) and power spectral density (PSD) and (2) area under the curve (AUC), are introduced to these classifiers. The algorithm is evaluated and validated with standard databases such as FANTASIA (healthy), MIT-BIH Arrhythmia (arrhythmic), and long-term ST database (ischemic), respectively. For uniform probability classification, ECG episodes with 100% sensitivity (Se) and the specificity (Sp) are included in this analytical modeling. As the experimentation is performed to validate the possibility of these features for classification, the percentage of classification certainly could be improved by considering other vital features. We conclude that correlative analysis of QRS-ST may be evoked as significant marker for arrhythmia and ischemia.