AI may spot deadly heart risk in a routine ECG

Key Points:

• UC Berkeley researchers developed an AI model trained on over 440,000 ECGs to identify hidden patterns linked to sudden cardiac death, outperforming current screening methods like left ventricular ejection fraction (LVEF) measurements.
• The AI detected a high-risk group with a 7% annual sudden cardiac death rate, higher than the 4.6% rate found using traditional LVEF criteria, and identified warning signs in routine ECGs that are typically missed by doctors.
• A specific ECG feature in the QRS complex, particularly in lead aVL, was highlighted by the AI as a novel predictor of sudden cardiac death, suggesting AI can uncover previously unknown risk indicators.
• This technology could improve decisions about implantable defibrillators by better identifying patients who need close monitoring, potentially reducing unnecessary procedures and saving lives through earlier intervention.
• While promising, the AI tool requires further validation across diverse populations and raises important privacy concerns about the use and control of patient data in medical AI development.