UC Berkeley AI model improves detection of sudden cardiac death risk

Researchers at the University of California, Berkeley have developed an artificial‑intelligence algorithm that identifies a previously unknown signal in electrocardiograms (ECG) to flag patients at high risk of sudden cardiac death. The model was trained on more than 440,000 ECGs from Sweden, linked with death certificates, and subsequently validated with anonymised ECG data from a San Diego hospital system in the United States and an independent dataset from Taipei, Taiwan.

The AI system outperformed standard clinical tests that measure cardiac output, isolating a high‑risk group with an annual mortality rate of 7 % compared with the 4.6 % rate identified by existing methods. The researchers estimate the approach could correctly identify thousands of additional patients each year who would otherwise be classified as low risk. The study, authored by associate professor Ziad Obermeyer and colleagues, was published in *Nature*.

Professor Obermeyer emphasised the clinical relevance, stating, "Taking medical decisions is really hard, and that's why AI is useful. It helps us make better choices and start to understand what really happens to these patients before their hearts stop."