On skeletonization of blood vessels in angtographic MRI images of human brain

Abstract

The aim of this paper is to investigate the main existing 3D skeletonization methods, their properties and particularities with respect to centerline extraction of segmented blood vessels in human brain MRI images. Skeletonization of 3D objects may be performed using different approaches. Basically there are four main concepts of skeletonization algorithms which may be adopted in automatic analysis. In the article we use four algorithms representing each concept and obtain results in the cases of modeled and real vessels extracted from MRI volume. We investigate the main properties of obtained skeletons and formulate essential and preferable properties of 3D skeleton curves. According to visual analysis the most suitable method was selected and motivated.