Integrating Markov model and morphology analysis for finer classification of ventricular arrhythmia in real time

In this paper, we propose a novel technique which integrates morphological analysis of ECG signal with Markov Model to derive four major subclasses of ventricular arrhythmia in real time. The subclasses are: ventricular tachycardia, ventricular fibrillation, ventricular flutter and premature ventricular complex. Markov Models have been trained using MIT-BIH ventricular arrhythmia database. Parameters of Markov Models are adapted in real time for various patients. The accuracy of classification has been further improved by detecting the P waves potentially embedded inside QRS complex using an area subtraction method. The technique has been tested on standard public database, and it diagnoses the subclasses with high accuracy in real time.