Adaptive frequency tracking on the ECG used to predict the success of electrical cardioversion of atrial fibrillation

One of the rhythm control strategies in atrial fibrillation management consists in the application of a brief single electrical shock to the atria. This electrical cardioversion does not always succeed, and the long-term success of this therapy is not sufficiently predictable on the basis of clinical and echocardiographic parameters. In noninvasive atrial fibrillation studies, the frequencies observed in the atrial ECG signals are considered as an indicator of the underlying dynamics. In this study, we examine the use of frequency tracking techniques in predicting the success of electrical cardioversion. A clinical database of 18 patients in sustained AF prior to an electrical cardioversion attempt was employed. The signal processing steps performed on 5-minute standard 12-lead ECG signals were: baseline correction, ventricular activity cancellation followed by an adaptive frequency tracking algorithm. Fisherpsilas linear discriminant function yielded up to 94% correct classifications (p=10^-3).