3D Cardiac Magnetic Resonance Substrate Features-based Machine-learning Model for Postmyocardial Infarction Risk...

Area of Science:

• Cardiovascular Imaging
• Machine Learning in Medicine
• Cardiac MRI

Background:

• Accurate risk stratification after myocardial infarction (MI) is crucial but challenging.
• Integrating advanced imaging features can enhance predictive capabilities.

Purpose of the Study:

• To develop interpretable machine learning (ML) models for predicting major adverse cardiovascular events (MACE) post-MI.
• To assess the value of 3D cardiac magnetic resonance (CMR) substrate features in risk stratification.

Main Methods:

• Retrospective analysis of 292 MI patients undergoing CMR.
• Extraction of 3D CMR substrate features (core scar, border zone, abnormal corridors).
• Development and validation of ML models, including TabPFN, with SHAP analysis for interpretability.

Main Results:

• Nine key predictors identified: clinical, functional, and 3D CMR substrate features.
• ML model combining clinical and 3D features achieved high performance (AUC 0.89 external).
• 3D CMR features alone (AUC 0.90 external) outperformed clinical and functional models.

Conclusions:

• ML models incorporating 3D CMR substrate features significantly enhance post-MI MACE prediction.
• These models provide interpretable and personalized risk stratification tools.
• 3D CMR substrate analysis offers a valuable addition to traditional risk assessment methods.