AI Model Outperforms Doctors at Identifying Patients Most At-Risk of Cardiac Arrest

Hypertrophic cardiomyopathy is one of the most common inherited heart conditions and a leading cause of sudden cardiac death in young individuals and athletes. While many patients live normal lives, some are at a significantly increased risk of fatal cardiac events. Identifying which patients are at high risk has long been a challenge, with current clinical guidelines in the U.S. and Europe proving to be only about 50% accurate, no better than a coin toss. This has led to critical under-protection for high-risk individuals and unnecessary implantation of defibrillators in others. A new approach that can better identify at-risk patients and reduce unneeded interventions has now been developed and tested with significantly higher accuracy.

Researchers at Johns Hopkins University (Baltimore, MD, USA) have developed a deep learning-based artificial intelligence model called Multimodal AI for Ventricular Arrhythmia Risk Stratification (MAARS). The system was designed to analyze a full range of patient medical records in combination with contrast-enhanced MRI images of the heart. Unlike traditional methods, MAARS can interpret complex scarring patterns—fibrosis—found in the hearts of patients with hypertrophic cardiomyopathy, which are known to raise the risk of sudden cardiac death. While clinicians have struggled to make sense of this imaging data, the AI model was able to extract and utilize hidden predictive information from the scans. By doing so, the model not only predicts a patient’s risk but also explains the reasoning behind the assessment, enabling doctors to tailor personalized care plans. MAARS builds upon prior work from the same team that developed an AI model in 2022 for predicting cardiac arrest in infarct patients.

The researchers tested the new model on real-world data from patients treated at Johns Hopkins Hospital and Sanger Heart & Vascular Institute. The study, published in Nature Cardiovascular Research, showed that the AI model achieved 89% accuracy overall and 93% accuracy for patients aged 40–60, the demographic at highest risk. The tool significantly outperformed current clinical guidelines across all demographics. Beyond improving survival prediction, the model offers transparency in its decision-making process, a feature that enhances clinical trust and utility. The team now plans to validate MAARS on larger populations and adapt it for use with other heart conditions, such as cardiac sarcoidosis and arrhythmogenic right ventricular cardiomyopathy.

“Currently we have patients dying in the prime of their life because they aren’t protected and others who are putting up with defibrillators for the rest of their lives with no benefit,” said senior author Natalia Trayanova, a researcher focused on using AI in cardiology. “We have the ability to predict with very high accuracy whether a patient is at very high risk for sudden cardiac death or not.”

Related Links:
Johns Hopkins Hospital and Sanger Heart & Vascular Institute