Digital Twins (DTs) of atrial fibrillation (AF) offer a promising framework for selecting personalized ablation strategies, but their calibration to patient-specific electrophysiological activity remains challenging.
In this work, ECGI-derived dominant frequency (DF) maps were used to compute cycle length (CL) distributions in 4 patients. These maps were integrated into patient-specific left atrial models and used to calibrate spatially heterogeneous electrophysiological remodeling. Personalized DTs were then simulated using a rapid pacing protocol to induce AF, and two ablation strategies (pulmonary veins isolation (PVI) and Dominant Frequency Adenosine Stability Index (DFASI-guided) were evaluated.
Simulated CL maps showed good agreement with ECGI-derived maps, with mean absolute errors ranging from 8 ms to 48 ms and 99% of nodes below 50 ms error. AF was induced and maintained in all patients, enabling the identification of arrhythmogenic regions. DFASI-guided ablation resulted in 0% inducibility of atrial arrhythmias, compared to 33% in the non-ablation scenario and 50% for PVI.
This study presents a methodology to calibrate AF DT using ECGI-derived biomarkers and evaluate ablation strategies in silico. The proposed framework enables non-invasive, patient-specific therapy planning and may improve the identification of optimal ablation targets.