An ECG biomarker for sudden cardiac death discovered with deep learning
Researchers used deep learning to discover an ECG-based biomarker that can identify individuals at high risk of sudden cardiac death, potentially enabling earlier prevention and intervention strategies.
Background
- Sudden cardiac death (SCD) kills roughly 1 in 10 people worldwide — a heart rhythm failure that strikes without warning, often in people who seemed healthy.
- Doctors have no reliable way to predict who is at risk. Standard tests (ejection fraction, ECG reading by cardiologists) miss most cases.
- The study used a deep-learning model (a type of AI) trained on nearly 2 million electrocardiograms (ECGs) from 570,000 patients to find a hidden signal in the heartbeat that predicts SCD years in advance.
- The model discovered a biomarker called "R-wing" — a subtle shape in the ECG's QRS complex that human doctors cannot reliably see — which outperforms all existing clinical risk scores.
- The authors validated the finding across multiple hospitals, countries, and ethnic groups, and even showed that the biomarker has a plausible biological basis (connected to calcium handling in heart cells).
- This is a landmark example of AI finding a genuinely new medical signal that humans had missed, not just automating what doctors already do.