Real-world experience with second-generation artificial intelligence algorithm software guidance for de novo and repeat catheter ablation of long-standing persistent atrial fibrillation patients

Abstract Background Despite the availability of advanced ablation strategies, the procedural workflow associated with signal-based ablation of persistent atrial fibrillation (AF) remains lengthy and intricate. Recently, the Tailored-AF trial using a first-generation artificial intelligence (AI) software-guided persistent AF ablation demonstrated superiority to a pulmonary vein isolation (PVI)-only procedure in arrhythmic outcome. Purpose To investigate the feasibility and safety of real-world usage of a second-generation multipolar electrogram analysis AI algorithm software allowing for real-time auto-tagging during de novo and repeat catheter ablation of long-standing persistent AF patients. Methods In this prospective, single-centre, non-randomized study, 100 consecutive long-standing persistent AF patients underwent catheter ablation. Concomitantly with electro-anatomical high-density anatomy and voltage acquisition, an automated tagging-enabled, second-generation software allowed for AI-guided spatiotemporal dispersion (DISPERS) mapping. After LA mapping, pulmonary veins (PV) were isolated or re-isolated in de novo and repeat ablation patients, respectively. Then, ablation of automatically-tagged DISPERS-areas was performed to homogenize, dissect, isolate, or connect DISPERS areas to non-conducting anatomical structures. Results Out of 100 patients (mean age 71±10 years, n=45 female), 56 patients were referred for a de novo procedure while 44 for a repeat procedure. Re-isolation of reconnected PV(s) was performed in 34/44 patients who received an AF ablation before first-time DISPERS-guided ablation. In 85/100 patients, at least one linear lesion was created (anterior line, n=39; roof line, n=38; floor line, n=13; cavotricuspid isthmus, n=4 and posterior box lesion, n=31). Average procedure time was 141±45 min with a fluoroscopy time of 11±8 min and a radiofrequency duration of 27±10 min. Following PVI, ablation of DISPERS areas led to AF cycle length prolongation (164±22 ms before ablation vs 190±27 ms after) and AF termination to atrial tachycardia (AT) or sinus rhythm (SR) in 22 patients. No stroke or pericardial tamponade were observed while groin complications occurred in 5 patients. Early recurrence (£42 days) was recorded in 32 patients (18 organized AT, 10 AF, 4 AF/AT). After a 6-week-blanking period, recurrence was documented in 8 patients. Re-ablation was performed in 6 AT patients. After a mean of 1.5 months (50±115 days), 92/100 patients remained in stable SR, of whom 32/92 required a period of antiarrhythmic drugs. Conclusion This study reports the first real-world experience with a second-generation AI algorithm software guidance for de novo and repeat catheter ablation of long-standing persistent AF patients. This technology allows for a quick, operator- and experience-independent annotation of purported AF drivers in the LA, followed with a signal-based and patient-specific ablation of persistent AF.PA/AP view of ablated LA dispersions