A review of vessel extraction techniques and algorithms
ACM Computing Surveys (CSUR)
Models, algorithms and applications in vascular image segmentation
Machine Graphics & Vision International Journal
Hi-index | 0.00 |
The article is dealing with the automated extraction of branching structures in 3D medical images. A generic object-oriented programming framework is proposed, in which most existing iterative algorithms for centerline extraction in tubular objects can be efficiently implemented, and the bifurcations can be handled. New algorithms can thus easily be derived. We describe a simple algorithm for fast extraction of the 3D structure of the vascular tree, which has been implemented within this framework. The algorithm recursively tracks the branches and detects the bifurcations by analyzing the binary connected components on the surface of a sphere that moves along the vessels. It assumes that the vessels can locally be separated from the background by an appropriate adaptive threshold. The originality of the algorithm resides in the analysis of the evolution of the connected components during the sphere growth that allows it to cope with local abrupt changes of the vessel diameter and shape. It was successfully tested in 16 magnetic resonance angiography images. Its accuracy was assessed by comparing the resulting axes with those extracted by a reference algorithm. The distance between them was less than one voxel except in bifurcations, where the maximum distance was 3.8 voxels.