Predicting initiation and termination of atrial fibrillation from the ECG

Atrial fibrillation is the most common cardiac arrhythmia, affecting more than two million people in the US. Several therapies for patients with atrial fibrillation are available, but methods to help physicians select the optimal therapy for an individual patient are still required. Knowledge of whether a patient with a normal ECG will exhibit atrial fibrillation in the future, as well as whether atrial fibrillation will terminate spontaneously, would be very useful in clinical routine. The paper presents a software system for predicting the initiation and termination of atrial fibrillation from the ECG. The algorithms have been validated on ECGs from several signal databases. Prediction of the initiation of atrial fibrillation was achieved by detecting premature heart beats and analyzing the morphology of their P waves. Prediction of the termination of atrial fibrillation was based on calculation of the major atrial frequency. This frequency has been shown to decrease significantly prior to the termination of atrial fibrillation. Nevertheless, the effect is much less distinct in the large data set used for this study compared to previous studies. The initiation of atrial fibrillation, however, could be correctly predicted in approximately 75% of the data analyzed.