An AI found the heartbeat clue that could save thousands from sudden cardiac death

Researchers at UC Berkeley have used artificial intelligence to discover a previously hidden warning signal in standard electrocardiogram (EKG) scans that predicts sudden cardiac death far more accurately than conventional methods. By training the system on over 440,000 EKGs from Sweden and validating it against patient records from the US and Taiwan, they identified subtle electrical patterns that indicate significantly higher risk, even in patients who appear low-risk under current medical guidelines.

This AI-powered analysis could fundamentally change how heart patients are screened worldwide. Unlike traditional methods that focus on how well the heart pumps blood, this technology works with existing, routine EKG images, making it a practical upgrade to standard care. It identified a group with a 7% annual death rate compared to the 4.6% rate from conventional tests, potentially allowing doctors to prevent thousands of otherwise unexpected fatalities by better targeting who needs implanted defibrillators. The breakthrough shows how AI can find life-saving patterns in data we've been collecting for decades but never understood.