SEP 4, 2026 · PREPRINT
Learning from VAE Errors to support ECG-based Differential Diagnosis of Myocardial Scar
arXiv
Single-center proof-of-concept study using machine learning on ECG to detect myocardial scar; modest discrimination (AUC 0.686 best) without peer review, establishing feasibility rather than clinical utility.
Reported
Best AUC (ECGx.AI + Random Forest)0.686
AUC (β-VAE + Gradient Boosting)0.577
Sensitivity (β-VAE + Gradient Boo…0.775