Simulation-Driven Machine Learning for Predicting Stent Expansion in Calcified Coronary Artery.

Area of Science:

• Biomedical Engineering
• Computational Mechanics
• Medical Imaging

Background:

• Coronary artery stenting is crucial for treating calcified lesions.
• Accurate prediction of stent expansion is vital for procedural success.
• Existing methods may not fully capture the complexities of stent deployment in calcified arteries.

Purpose of the Study:

• To develop and validate a simulation-driven machine learning framework for predicting stent expansion.
• To integrate finite element (FE) analysis with machine learning (ML) models.
• To assess the predictive capabilities of different ML models and feature sets.

Main Methods:

• Patient-specific coronary artery models were reconstructed from optical coherence tomography (OCT) images.
• Finite element (FE) simulations captured the stenting procedure.
• Geometric features were extracted from pre-stenting models for training ML models (linear regression, support vector regression).
• Stretch and calcification features were analyzed for their predictive power.

Main Results:

• Support vector regression (SVR) models demonstrated superior prediction accuracy over linear regression, with lower bias.
• Incorporating stretch features, based on mechanistic understanding, improved prediction compared to calcification features alone.
• Averaging features over neighboring cross-sections did not significantly alter prediction bias or error range.

Conclusions:

• The developed simulation-driven ML framework enhances mechanistic understanding of stenting in calcified coronary arteries.
• This approach shows promise for precise prediction of stent expansion.
• The study highlights the importance of biomechanical features in predicting stenting outcomes.