Cardiovascular Tree Recognition and its Three-Dimensional Reconstruction from Monoplane X-Ray Angiography Based on Spatio-Temporal Constraints

In this paper, we propose a new method of collaborative matching based on spatio-temporal constraints (COMSTC) that can be used for three-mensional (3D) reconstruction of monoplane X-ray coronary angiography. The main idea of this method is that, under the guidance of the multi-scale sptio-temporal expectation model (MSSTEM), the constraint relationships such as the time of the contrast agent flow, the spatial distribution of vasculum primitives, anatomical structure, and topological similarity are collaboratively used to distinguish the homonymous vasculum primitive pairs of dual-perspective angiogram pairs of the sequence. As can be seen from the experimental results, the MSSTEM of the vascular tree in this paper is used to guide matching of the blood vessels systematically starting from the root, which solves the problem of individual diversity, and the controllability and robustness are satisfactory. Iterative matching and searching using the maximum potential descent method improve the matching accuracies of pairs of vasculum primitives and pairs of points on vasculum primitives, and guarantee a more reliable automatic 3D reconstruction of the dual-perspective angiography images.