A hidden predictor of sudden cardiac death uncovered by deep learning

Researchers developed a deep learning model that predicts sudden cardiac death by analyzing electrocardiogram (ECG) recordings, as reported in Nature on June 24, 2026. The AI model, trained on a dataset of over 100,000 ECGs, identified subtle patterns that are not apparent to human interpretation. These patterns are linked to a specific type of electrical instability in the heart muscle, known as transmural dispersion of repolarization. The study, led by researchers at the University of Oxford, demonstrated that the model could identify individuals at high risk of sudden cardiac death up to 10 years in advance. This predictive capability surpasses existing risk stratification methods, which often rely on factors like age, sex, and clinical history. The model achieved an area under the receiver operating characteristic curve (AUC) of 0.85 in predicting cardiac events within a 5-year window. The findings suggest that this AI-driven approach could revolutionize how sudden cardiac death risk is assessed, potentially leading to earlier interventions and improved patient outcomes. Further validation studies are planned to integrate this technology into clinical practice.