A computationally light-weight real-time classification method to identify different ECG signals

Abstract

Ventricular arrhythmia is the main cause of cardiac arrest in patients with chronic heart disease. An undetected episode of ventricular tachycardia (VT) can be fatal if emergency medical assistance is not provided. Therefore, it is important to devise a real-time mobile ECG signal analysis algorithm for detection of ventricular tachycardia (VT). This paper presents an algorithm for automatic identification of normal sinus rhythm (NSR) and ventricular tachycardia (VT) which is applicable in a mobile environment. The algorithm employs peak-valley detector and cross-correlation technique to compute a feature vector. The selected features are beats-per-minute (BPM), NSR template accuracy and VT template accuracy. Based on the selected features, a fuzzy k-NN classifier is trained for classification. The algorithm specificity and sensitivity for classifying between NSR and VT ECG signal is 92.5% and 93.5% respectively.