Adaptive Multiple Frequency Tracking Algorithm: Detection of Stable Atrial Fibrillation Sources from Standard 12-Lead ECG
The detection of stable atrial fibrillation (AF) sources remains one of the major challenges in the AF management. In this study, we investigated the feasibility of detecting stable AF sources from (non invasive) data simulated by means of numerical procedures. By using a 3D biophysical model of the atria (Courtemanche membrane kinetics) and a compartmental torso model, 21 different episodes of AF were generated (transmembrane potentials through out the tissue and 12-lead ECGs). The stability of the episodes was established by visual inspection of the electrical propagation over the epicardial surface (group A: without a stable source, group B: with stable sources). This evaluation constitutes our gold standard. We hypothesized that during AF sustained by stable sources ECG signals include significant components at the frequencies related to the cycle length of these respective AF sources. These frequency components were jointly estimated on the 12-lead ECGs with an adaptive multiple frequency tracking algorithm. The ratio between the sum of the estimated frequency component powers and the sum of the 12-lead ECG signal powers was used as the discrimination feature r to estimate the number of sources. Nine simulated AF episodes were characterized by complex dynamics (group A). Group B comprised 11 simulated AF episodes having a single stable source and one having two stable sources. The r values observed were: group A, r = 0.05 ± 0.04 (mean ± SD) and group B, r = 0.28 ± 0.17. With a discrimination feature threshold set at 0.14, no stable AF source was detected in group A. Eight single AF sources were detected among the 11 ones of group B. The case with two AF sources in group B was also correctly classified. This corresponds to 85.7% correct classification, 100% sensitivity and 75% specificity. The proposed approach provides information about the presence stable AF sources. This information may lead to a more accurate identification of patients suitable for specific AF ablation procedures.