Novel AI-based physiological tracking of waves in atrial fibrillation identify active pulmonary veins, non-PV sites of termination

Background A major challenge in ablating atrial fibrillation (AF) is to efficiently identify important sites outside the pulmonary veins, particularly in patients with persistent AF. An often observed yet unexplained phenomenon is transient organization of waves during AF (Fig. A). Hypothesis: We hypothesized that mapping AF waves for repeating patterns (fig. A), using novel artificial-intelligence (AI) tools to accurately identify local activity from noisy electrograms, could identify patients with active pulmonary veins (PV), where AF waves predominantly exit (versus enter) PVs, or emanate from non-PV sites which acutely maintain AF. Methods We developed an AI-system to identify repeating AF waves on multipolar catheters (fig B), trained first to separate local activity from far-field activity from >10 million AF electrograms then second, on a multielectrode array, to aggregate predominant propagation directions in N=174 AF patients (68.2±8.7 years old, 13.7% females, 72.6% persistent AF). The AI-system was used to track AF waves near the PVs and to track AF waves near and remote from sites of acute termination by ablation in a validation cohort (26,640 electrograms, N=62). Results In Fig. C, AF wave propagation near and remote from PVs was linked to clinical endpoints. AF waves were variable yet with repetitive activity in >75% of patients. Near the PVs, Fig. C shows waves exiting PVs (i.e. active PVs) in a 60 Y man with persistent AF and successful ablation. The inset shows that, overall, PVI-based AF ablation was more effective in patients with active versus passive PVs. Fig. D shows AF in a 72Y woman with persistent AF, with AF waves predominantly exiting a non-PV site of AF termination in the right atrium. Overall, wave vectors exited termination sites significantly more often than random control sites (p<0.001). Conclusions A novel AI-based system to track AF waves trained using millions of electrograms showed that PVI- is more effective in patients with active pulmonary veins, while waves exiting non-PV sites may indicate sites of AF termination. Prospective studies of these approaches are warranted.