Evaluation of AI-determined dispersion patterns in spontaneous versus induced atrial fibrillation among persistent AF patients

Abstract Introduction The optimal ablation strategy for persistent AF (pers AF) is still a matter of research. Findings from the recent TAILORED-AF trial have shown that a tailored ablation approach—specifically targeting AI-identified spatiotemporal dispersion areas in addition to PVI—achieves superior outcomes compared to PVI alone. It is unclear, though, whether this approach is effective for patients in sinus rhythm upon admission post-cardioversion (CV), for whom AF would need to be induced prior to ablation. One key assumption is that the spatiotemporal dispersion in spontaneous AF (spont-AF) and induced AF (ind-AF) is comparably distributed. Objective Our study sought to evaluate and compare the dispersion characteristics (location and extent) detected by the Volta AF-Xplorer software in spont-AF and ind-AF settings. Methods Eligible patients included those with persAF who were scheduled for a tailored ablation procedure combined with PVI. In our clinical practice, we first create a left atrial (LA) electroanatomic spont-AF map, followed by electrical CV and creation of a second map during atrial pacing. If atrial pacing incidentally triggered AF, an ind-AF map was generated . The LA was divided into 16 standardized subregions, and each area’s dispersion characteristics were classified based on AI assessment. Two types of analyses were performed: (1) a dichotomous analysis to determine if spatiotemporal dispersion was present in spont-AF versus ind-AF, and (2) a semi-quantitative assessment where subregions were categorized by dispersion levels of 0%, <10%, >10–50%, or >50%. Results From April to November 2024, 63 patients with persAF were treated with ablation at our center, among whom 10 patients (mean age 78±9 years, 50% female, 40% with long-standing persAF) had ind-AF maps generated. AF termination was achieved in 70%. Across both spont-AF and ind-AF, 160 segments were mapped and assessed for dispersion. Spatiotemporal dispersion was identified in 98 of the 160 spont-AF subregions and in 97 of the 160 ind-AF subregions (p=0.909). Only 5 out of 160 subregions (3%) differed in characterization between spont-AF and ind-AF maps. Semi-quantitatively, the subregion dispersion breakdown was as follows: spont-AF: 0% - 62 subregions, <10% - 45, >10–50% - 33, >50% - 20; ind-AF: 0% - 63, <10% - 48, >10–50% - 28, >50% - 21; p=0.817. In 20 cases (12.5%), a one-category variation was observed between spont-AF and ind-AF maps, while the classification remained consistent in the remaining 140 cases (87.5%), with no discrepancies greater than one category. Conclusion Our findings suggest that AI-adjudicated dispersion extent and distribution are consistent between spontaneous AF and induced AF. These results support the potential role of inducing AF prior to ablation to refine ablation strategies. Further studies are required to explore whether the TAILORED-AF trial findings are applicable to persAF patients requiring AF induction prior to ablation.