Coronary Artery 3D/2D Registration Based on Particle Swarm Optimization of Contextual Morphological Features

This paper proposes a novel method for coronary artery registration of computed tomography angiography and digital subtraction angiography (DSA) images. Firstly, the vascular centerlines of DSA are extracted, the sub-pixel corner points of the vessel centerline are identified by Shi-Tomas corner detection algorithm and gray central moment model. Then, the 2D vessel intersections are extracted by combining the skeleton erosion algorithm and the contextual morphological feature descriptors. For the extraction of 3D vessel intersections, we extract the 3D vessel centerline through the axial section centroid fitting algorithm, the 3D vessel intersections are extracted according to the centerline intersecting model. Finally, the pose of the 3D vessel model is searched by the particle swarm optimization algorithm based on the acquired 2D and 3D vessel intersections. The proposed method was validated on 10 clinical coronary cases, the accuracy of the proposed method is 0.25 ± 0.09 mm for the corner detection of 2D vessel centerlines, and 0.56 ± 0.11 mm for the extraction of 3D vessel centerline. The final 3D and 2D vessel registration accuracy is 2.28 ± 0.21 mm, which is better than the comparison methods.