Robust Centerline Extraction for Tree-like Blood Vessels based on the Region Growing Algorithm and Level-Set Method

Jing Xu, Daming Feng, Jian Wu, Zhiming Cui

Abstract: In this paper, we present a tree-like blood vessels based on the region growing algorithm and level sets method for computing centerlines for both 2D and 3D shape analysis. Consider the tree-like vascular structure, we first use region growing technology to generate the shortest distance from the given root to all the pixels, ground on the distance can detect all the extreme nodes of branches automatically. Propagate wave-front from a point of skeleton line with the level sets, track the points of the largest curvature on the surface of the wave-front, through solute differential equations, the skeleton can be gained. The proposed method is computationally inexpensive, not only maintain the position exactly and the shape smoothly and connectedly, and less sensitive to boundary noise.