Source Localization of Atrial Fibrillation using Probabilistic Dominant Frequency and Spatiotemporal Activation Time

Atrial fibrillation (AF) is a common cardiac arrhythmia with high prevalence and morbidity. AF can be managed by ablation therapy, where AF driving sources are identified and ablated to restore sinus rhythm. In this paper, we present a non-invasive approach to derive an improved metric for AF source localization using inverse reconstruction of atrial activity from body surface potential signals. We propose a combined metric involving the probabilistic distribution of dominant frequency and spatiotemporal activation time to identify atrial regions responsible for AF continuance. AF source localization results were reproduced for a clinical Electrocardiographic Imaging (ECGI)-AF data set and simulated atrial activity data. The generated results were in close accordance with ground truth. The proposed method aims at better source localization that can aid in ablation procedures by avoiding unnecessary interseptal punctures, reducing risk and procedural time.