Machine Learning for Coronary Plaque Characterization: A Multimodal Review of OCT, IVUS, and CCTA

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Computational Pathology

Background:

• Coronary plaque vulnerability, not just artery narrowing, is the primary cause of acute coronary syndromes.
• Current in vivo plaque imaging methods like optical coherence tomography (OCT), intravascular ultrasound (IVUS), and coronary computed tomography angiography (CCTA) require manual analysis, which is slow and subjective.
• Automated plaque analysis is needed to overcome the limitations of manual interpretation.

Purpose of the Study:

• To survey artificial intelligence (AI) applications, specifically machine learning (ML) architectures, for automated coronary plaque segmentation and risk characterization.
• To evaluate the performance of AI in identifying vulnerable plaque features across different imaging modalities (OCT, IVUS, CCTA).
• To assess the potential of AI-enhanced plaque assessment in clinical practice.

Main Methods:

• A narrative literature survey was conducted focusing on AI and ML applications in coronary plaque analysis.
• Reviewed studies utilized ML models for automated segmentation of lumen and plaque from OCT, IVUS, and CCTA data.
• Examined the ability of AI models to detect specific plaque vulnerability features and their impact on prognostic stratification.

Main Results:

• Recent ML models demonstrate expert-level performance in lumen and plaque segmentation.
• AI reliably detects key vulnerable plaque features, including lipid-rich necrotic core, calcification, positive remodeling, and the napkin-ring sign.
• Integrative radiomic and multimodal AI frameworks show promise in improving prediction of major adverse cardiac events.

Conclusions:

• AI-enhanced plaque assessment provides rapid, reproducible, and comprehensive analysis of coronary imaging.
• Current limitations include small datasets, varied validation metrics, and poor model interpretability.
• Future research should focus on large multicenter datasets, explainable AI, and prospective outcome validation for clinical integration.