SUDDEN cardiac death (SCD) remains a leading global cause of mortality, particularly driven by arrhythmic events. While implantable cardioverter defibrillators (ICDs) offer protection for high-risk individuals, accurately identifying who benefits most, especially among patients with hypertrophic cardiomyopathy (HCM), has been a longstanding clinical challenge.
HCM, the most common inherited heart disease, affects 1 in 200–500 people and is a primary cause of SCD in younger populations. Traditional risk stratification methods, such as using left ventricular ejection fraction (LVEF), often fall short in HCM, where patients typically retain normal or hyperdynamic heart function despite being at risk.
In a groundbreaking development, researchers have introduced MAARS (Multimodal AI for ventricular Arrhythmia Risk Stratification), an artificial intelligence model designed to predict the risk of SCD due to arrhythmias (SCDA) using diverse medical data. MAARS uniquely integrates electronic health records, imaging reports, and contrast-enhanced cardiac MRI using a transformer-based neural network, significantly outperforming existing clinical tools in predictive accuracy.
Tested across two patient cohorts from distinct healthcare systems, MAARS demonstrated impressive generalisability, fairness across demographic groups, and transparency in its decision-making process. A key strength lies in its use of raw MRI imaging data analysed through a 3D Vision Transformer (3D-ViT), allowing the model to detect subtle, disease-related changes in heart structure often overlooked by traditional assessments.
Beyond individual risk prediction, MAARS provides interpretable insights into population-level trends, reaffirming known risk factors while highlighting novel associations—such as the elevated SCDA risk in patients with non-obstructive HCM.
Although limitations remain, including relatively small patient cohorts and data-intensive requirements, MAARS marks a major advance in precision cardiology. With further validation, this AI-powered tool could transform how clinicians assess SCD risk in HCM patients—enabling more personalised, accurate, and timely interventions.
Reference
Lai C et al. Multimodal AI to forecast arrhythmic death in hypertrophic cardiomyopathy. Nat Cardiovasc Res. 2025;DOI:10.1038/s44161-025-00679-1
