Vascular Tree Matching from Multiple View Projections

Abstract

Generally, it is difficult to visually inspect the stenos is from angiographic projections during the diagnosis of coronary artery disease. However, 3D reconstruction methods can effectively remedy the shortage of angiographic imaging technique, which can recover coronary arteries in the 3D space from limited projections. For the 3D reconstruction technique, correspondence matching is the key to determine the accuracy of structure recovery. In this paper, a novel method is developed for the calculating of the correspondences from multiple view projections. Epipolar geometry constraint is commonly used for the correspondences matching. However, the epipolar line is usually intersected with multiple vessels through the epipolar geometry constraint. In order to solve this problem, in this paper, the curve matching method is introduced to estimate the geometry constraint, and hence to find the vascular tree matching. From our experiments, it can be seen that the developed method is very effective for the matching of vascular trees from multiple views.