Simple cone beam backprojection reconstruction for robust skeletonization of 3D vascular trees

Abstract

We describe and demonstrate a semi-automated method for three-dimensional skeletonization of vascular trees in X-ray CT image volumes. The key to the method, which tracks the vessel midline in successive increments, is that estimates of midpoint coordinates at each point along the vessel are obtained from a local reconstruction of the orthogonal vessel cross section at that point. Unlike filtered backprojection reconstruction, which attempts to recover the uniform density of a contrast agent-filled vessel across the entire lumen, simple backprojection reconstruction results in the minimum intensity occurring at the center of gravity of the orthogonal cross section. Our method is shown to provide a robust, semi-automated means for skeletonization of complex vascular trees.