AI and the World's Hard Problems · Medicine & drug discovery · Nature (computational biology) · 2026-10-25
Deep-learning-based electrode action potential mapping from annotation-free unipolar electrograms for substrate characterization in atrial fibrillation
Catheter ablation for non-paroxysmal atrial fibrillation is limited by difficulty identifying abnormal conduction regions. The team developed DEAP Mapping, a deep-learning method that reconstructs membrane potential from annotation-free unipolar electrograms. In ex vivo porcine hearts it matched optical mapping much more closely than conventional activation maps (SSIM 0.87 vs. 0.36), and in a retrospective study of 12 patients its conduction patterns partly differed from existing methods.