AI predicts patients likely to die of sudden cardiac arrest

Researchers in the Trayanova Lab at Johns Hopkins have developed an AI tool, MAARS (Multimodal AI for Arrhythmic Risk Stratification), that significantly improves the prediction of sudden cardiac death in patients with hypertrophic cardiomyopathy. By analyzing contrast-enhanced cardiac MRI scans alongside electronic health records, MAARS can detect subtle scar patterns in the heart and identify patients at high risk for fatal arrhythmias. In clinical testing, it achieved an accuracy of 89% overall and 93% for patients aged 40–60, outperforming current guideline-based methods.

The model not only predicts risk but also provides interpretable results that help clinicians understand why a patient is flagged as high risk—enabling more personalized treatment decisions. MAARS has the potential to reduce unnecessary defibrillator implants and ensure that those at true risk receive timely intervention. The research, funded by the NIH and published in Nature Cardiovascular Research, is now expanding to other heart conditions, with further clinical trials underway.