Skeletonization algorithm of high resolution vascular data

Abstract

Analysis of vascular structures can greatly improve accuracy of morphometric studies that may be used for early medical diagnosis. Laws and processes in vascular tree can be described by skeletal structure. Currently, there are numerous skeletonization methods that focus on computational efficiency. However, when data are too large to be loaded into the memory of a personal computer, such approaches are difficult to use. This paper presents an algorithm for constructing of skeleton, topologically equivalent to the original object, which processes data locally and at the same time preserves global attribute of the object - the centeredness. Built skeleton is based on distance maps and then can be used to calculate the morphological and topological characteristics.