Multimodal AI to forecast arrhythmic death in hypertrophic cardiomyopathy

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• Sudden cardiac death from ventricular arrhythmias is a major global health concern.
• Prognosticating arrhythmic death in hypertrophic cardiomyopathy (HCM) patients is difficult, with current guidelines showing limited accuracy.
• Existing risk stratification methods for HCM patients need improvement.

Purpose of the Study:

• To develop and validate a deep learning model, MAARS (Multimodal Artificial intelligence for ventricular Arrhythmia Risk Stratification), for predicting lethal ventricular arrhythmias in HCM.
• To improve the accuracy and fairness of risk stratification for HCM patients at risk of sudden cardiac death.

Main Methods:

• MAARS utilizes transformer-based neural networks to analyze multimodal data including electronic health records, echocardiogram and radiology reports, and contrast-enhanced cardiac magnetic resonance imaging.
• The model was trained and validated on internal and external patient cohorts.

Main Results:

• MAARS achieved an area under the curve (AUC) of 0.89 in the internal cohort and 0.81 in the external cohort.
• The model significantly outperformed current clinical guidelines, with AUC improvements of 0.27-0.35 (internal) and 0.22-0.30 (external).
• MAARS demonstrated fairness across demographic subgroups, unlike current guidelines.

Conclusions:

• MAARS represents a novel and effective deep learning approach for predicting lethal ventricular arrhythmias in HCM patients.
• The model's ability to integrate multimodal data, including cardiac MRI, enhances predictive accuracy.
• MAARS offers a more accurate, fair, and transparent tool for risk stratification in HCM, potentially improving patient outcomes.