PPCAR-Net: Projection-Refined Parametric 3D Coronary Artery Reconstruction from Sparse X-ray Angiographic Views

2026-10-06

AI Summary

This study addresses the challenges of vessel overlap matching and the inability of volumetric prediction to directly yield centerlines and radii in sparse-view 3D coronary artery reconstruction. To this end, we propose a projection-refined parameterized network that introduces a novel triangulation-free parametric representation. By integrating frozen VGGT features with branch queries within a coarse-to-fine inference framework, the method employs B-spline modeling alongside a projection-guided geometry-radius residual refiner to directly predict vascular topology without explicit point matching or intermediate volumetric representations. Experimental results on simulated angiographic data demonstrate that our approach significantly improves the reconstruction accuracy and connectivity of the right coronary artery. Furthermore, achieving single-inference latency of only 121 milliseconds, the proposed method exhibits strong potential for real-time clinical applications.

Abstract

Sparse-view 3D coronary reconstruction commonly relies on cross-view correspondence and triangulation, which are vulnerable to vessel overlap and foreshortening, or on volumetric prediction followed by vascular-graph extraction, which does not directly provide centrelines and radii. We introduce PPCAR-Net, a projection-refined parametric coronary artery reconstruction network that directly predicts a branch-structured centreline-and-radius representation without explicit point matching, triangulation, or an intermediate volume. Given a variable number of segmented views, a coarse predictor combines frozen VGGT features with learned branch queries to estimate branch presence, B-spline centreline trajectories, and dense radius profiles. Projection-guided geometry and radius refiners then sample local evidence from the input views and apply residual corrections learned with 3D supervision. We evaluate representation fidelity and sparse-view reconstruction quantitatively and qualitatively. On simulated angiographic masks generated from CT-derived coronary anatomy, PPCAR-Net produces better connected artery reconstructions and achieves strong centreline accuracy, particularly for RCA, while maintaining competitive volumetric overlap. Coarse-to-fine inference takes 121 ms, enabling real-time reconstruction.

Problem

• Sparse-view 3D reconstruction
• Coronary artery
• Centreline and radius representation
• X-ray angiography

Innovation

• Sparse-view 3D reconstruction
• Parametric coronary artery representation
• Projection-guided refinement
• B-spline centreline
• Real-time inference