A dual-channel deep learning framework integrating preoperative and intraoperative multi-phase data for predicting post-ablation recurrence in atrial fibrillation

Abstract

Predicting atrial fibrillation (AF) recurrence after catheter ablation is essential for optimizing postoperative management. This study developed the Ablation-Clinical Fusion Transformer (ACFT), a dual-channel framework integrating structured intraoperative ablation sequences with clinical indicators using Transformer and Multilayer Perceptron (MLP) architectures. Validated on real-world data, the model achieved an Area Under the Receiver Operating Characteristic curve of 0.844 (95% CI: 0.819, 0.869), representing a 7% improvement over standard benchmarks. This work provides an innovative approach for predicting AF recurrence after ablation and demonstrates significant potential for clinical application.