Leveraging Diffusion Model and Image Foundation Model for Improved Correspondence Matching in Coronary Angiography

2025-03-31

AI Summary

Cross-view correspondence matching in coronary angiography images is challenging due to texture scarcity, low contrast, and structural occlusion. Method: We propose a medical-prior-guided paradigm for paired image synthesis and semantic matching. Specifically, we introduce a conditional diffusion model guided by 3D CCTA mesh projections to synthesize high-fidelity paired X-ray images; further, we design a keypoint-aware semantic feature aggregation network leveraging vision foundation models (e.g., DINOv2 and SAM). Contribution/Results: Theoretically, we characterize the adaptation principles of vision foundation models for sparse, low-quality X-ray matching. Experimentally, our method achieves an 18.7% improvement in matching accuracy on synthetic data and demonstrates strong cross-domain generalization to real clinical data—outperforming prior methods significantly. This work establishes a practical, sample-efficient framework for medical image correspondence matching.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Calibration & Uncertainty Quantification
• Constraint Satisfaction and Optimization: Other Foundations of Constraint Satisfaction

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Accurate correspondence matching in coronary angiography images is crucial for reconstructing 3D coronary artery structures, which is essential for precise diagnosis and treatment planning of coronary artery disease (CAD). Traditional matching methods for natural images often fail to generalize to X-ray images due to inherent differences such as lack of texture, lower contrast, and overlapping structures, compounded by insufficient training data. To address these challenges, we propose a novel pipeline that generates realistic paired coronary angiography images using a diffusion model conditioned on 2D projections of 3D reconstructed meshes from Coronary Computed Tomography Angiography (CCTA), providing high-quality synthetic data for training. Additionally, we employ large-scale image foundation models to guide feature aggregation, enhancing correspondence matching accuracy by focusing on semantically relevant regions and keypoints. Our approach demonstrates superior matching performance on synthetic datasets and effectively generalizes to real-world datasets, offering a practical solution for this task. Furthermore, our work investigates the efficacy of different foundation models in correspondence matching, providing novel insights into leveraging advanced image foundation models for medical imaging applications.

Problem

• Improving correspondence matching in coronary angiography images
• Generating realistic synthetic data for training using diffusion models
• Enhancing matching accuracy with image foundation models

Innovation

• Diffusion model generates synthetic coronary angiography images
• Image foundation models guide feature aggregation
• Enhanced matching accuracy via semantically relevant regions