AI-Driven Voltage Map Analysis for Optimising Catheter Ablation Strategy in Atrial Fibrillation: A Proof-of-Concept Study

Abstract Background Pulmonary vein isolation (PVI) has been established as the standard catheter ablation (CA) strategy for atrial fibrillation (AF). However, approximately 20–40% of patients experience recurrence after CA. Although three-dimensional (3D) maps generated during CA provide valuable electrophysiological information, they may not be fully utilised in clinical decision-making. Objectives To develop an artificial intelligence (AI) model that analyses 3D voltage maps and long-term AF recurrence to guide best practices in CA for AF. Methods A dedicated multicentre registry recording detailed CA data for AF and recurrence was used to develop the AI model. The model was designed to evaluate the completion of PVI and ablation beyond-PVI (be-PVI), considering future AF recurrence with the need for additional PVI and be-PVI interventions. Results The AI model was trained and validated via fivefold cross-validation with 1,268 maps. It effectively stratified cases for predicting one-year AF recurrence after CA (p<0.001) and identified those likely to benefit from additional ablations (PVI: p = 0.032, be-PVI: p<0.001, and a combination of PVI and be-PVI: p<0.001). Conclusions The developed AI model predicts AF recurrence based on the completion of PVI and be-PVI and accurately identifies patients who may require further intervention. AI analysis of intraoperative 3D maps could guide optimal CA strategy planning, considering long-term AF recurrence.