A multicenter study reports an explainable machine learning model that combines cardiac CT angiography features with clinical data to predict which patients with atrial fibrillation and heart failure will improve after ablation. The approach aims to support patient selection rather than replace procedural decision-making. The key technical angle is interpretability, with the model designed to show which input signals drive predictions—an important requirement for clinical adoption and for aligning AI outputs with physician reasoning. For cardio-oncology and device partners, the result adds momentum to AI risk stratification tied to interventional outcomes.