Adaptive multiple frequency tracking algorithm: Detection of stable atrial fibrillation sources from standard 12-lead ECG

In this study we investigate a means of distinguishing between stable and more complex atrial fibrillation (AF) sources by analyzing ECG signals. For this purpose, 21 episodes of AF were generated by using a 3D biophysical model of the atria. The AF episodes were classified into two groups (with or without stable sources) by visual observation of the electrical propagation on the epicardial tissue (gold standard). The simulated 12-lead ECGs of these AF episodes were computed by using a compartmental torso model. The analysis of the ECG signals was performed by applying an adaptive multiple frequency tracking algorithm. The normalized power outputs of the algorithm directly provided information concerning the stability level. The comparison of the results of our method with the gold standard yielded 85.7% of correct classifications, with a sensitivity of 100% and 75% specificity.