A Cyclostationary Least Mean Squares Algorithm For Discrimination Of Ventricular Tachycardia From Sinus Rhythm

Abstract

A simple real-time algorithm is presented which accurately predicts a single beat of either sinus rhythm (SR) or ventricular tachycardia (VT) and detects a change from one rhythm to the other. The proposed algorithm learns from previous beats, so the prediction error decreases with time, thus accurately predicting the electrocardiogram. The algorithm detects changes in morphology (such as occurs during the spontaneous onset of VT during SR) by monitoring sudden increases in prediction error and is successful in all cases.