
AI Finds Hidden ECG Clue That Could Signal Sudden Cardiac Death Early
UC Berkeley researchers trained an AI on more than 440,000 ECGs and validated it with data from Sweden, the U.S. and Taiwan, uncovering a previously hidden ECG pattern linked to sudden cardiac death even when standard screenings look normal. The AI identified a high‑risk group with about 7% annual SCD risk versus 4.6% in the conventional reduced ejection fraction group, and many cases would have been missed by LVEF alone. If further testing confirms effectiveness across more populations, the approach could guide closer monitoring and influence defibrillator decisions, though it isn’t ready for patient use and raises privacy/data governance questions.









