Detecting intersection zones in thread-like 3D voxel clouds: a novel algorithm based on neighbourhood tracing

Indexed in SCOPUS - 3D topological segmentation entails the partition of 3D clouds of connected voxels (e.g. 3D images) into sets of entities, connected by a series of nodes. Especially for thread-like structures (e.g. blood vessels, proteins, piping and voxel-based 3D sketches), nodes detection and, more in particular, identification of entities converging in a node may be considered a crucial issue to correctly interpret the structure" of the analysed object. The main objective of the present work is to describe an innovative algorithm able to determine the possible intersection zones between groups of voxels composing a thread-like 3D voxel cloud. The devised algorithm is based on a multi-directional ray-tracing procedure applied, for each voxel of the 3D dataset, with reference to a set of directions defined by an appropriate 3D neighbourhood domain. The thread-like structure thicknesses, coming from the ray-trace analysis, are estimated along different directions. Finally, the set of thicknesses is analysed by means of a statistical approach so that voxels, possibly belonging to intersection zones, can be identified. The proposed algorithm demonstrated its effectiveness in a number of test cases.