NeRF-CA: Dynamic Reconstruction of X-ray Coronary Angiography with Extremely Sparse-views

2024-08-29

arXiv.org

AI Summary

Reconstructing 4D dynamic 3D coronary artery (CA) structures from sparse-view (only four frames) 2D X-ray coronary angiography (CA) sequences remains challenging due to severe cardiac motion, background clutter, and extreme data scarcity. Method: We propose an unsupervised, motion-guided Neural Radiance Fields (NeRF)-based reconstruction framework. For the first time, we explicitly model coronary dynamic motion within scene decomposition, jointly enforcing dynamic structural sparsity and scene smoothness regularization to precisely decouple the moving coronary foreground from static anatomical background. Our method integrates motion-guided decomposition, foreground–background disentanglement, and sparse-view optimization. Contribution/Results: On a 4D synthetic CA dataset, our approach quantitatively outperforms existing sparse-view X-ray NeRF methods. Qualitative evaluation demonstrates robust preservation of vascular continuity and temporal motion consistency even under ultra-low-dose conditions (four views), indicating strong clinical feasibility.

Technology Category

• Computer Vision: Motion & Tracking
• Machine Learning: Multi-instance/Multi-view Learning
• Reasoning under Uncertainty: Sequential Decision Making

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

Dynamic three-dimensional (4D) reconstruction from two-dimensional X-ray coronary angiography (CA) remains a significant clinical problem. Challenges include sparse-view settings, intra-scan motion, and complex vessel morphology such as structure sparsity and background occlusion. Existing CA reconstruction methods often require extensive user interaction or large training datasets. On the other hand, Neural Radiance Field (NeRF), a promising deep learning technique, has successfully reconstructed high-fidelity static scenes for natural and medical scenes. Recent work, however, identified that sparse-views, background occlusion, and dynamics still pose a challenge when applying NeRF in the X-ray angiography context. Meanwhile, many successful works for natural scenes propose regularization for sparse-view reconstruction or scene decomposition to handle dynamics. However, these techniques do not directly translate to the CA context, where both challenges and background occlusion are significant. This paper introduces NeRF-CA, the first step toward a 4D CA reconstruction method that achieves reconstructions from sparse coronary angiograms with cardiac motion. We leverage the motion of the coronary artery to decouple the scene into a dynamic coronary artery component and static background. We combine this scene decomposition with tailored regularization techniques. These techniques enforce the separation of the coronary artery from the background by enforcing dynamic structure sparsity and scene smoothness. By uniquely combining these approaches, we achieve 4D reconstructions from as few as four angiogram sequences. This setting aligns with clinical workflows while outperforming state-of-the-art X-ray sparse-view NeRF reconstruction techniques. We validate our approach quantitatively and qualitatively using 4D phantom datasets and ablation studies.

Problem

• Dynamic 4D reconstruction from 2D X-ray angiography with sparse views
• Overcoming sparse-views and cardiac motion in coronary artery imaging
• Achieving automatic reconstruction without extensive user input or large datasets

Innovation

• NeRF-CA decouples moving artery and static background
• Achieves 4D reconstruction from four angiograms
• Outperforms sparse-view X-ray NeRF methods