Segmentation and Symbolic Description of Cerebral Vessel Structure based on MR Angiography Volume Data

Therapy and planning require more information from digital imagery than simply the presence of disease, since the medical task being the measurement and identification of structures. The present paper focuses on the conversion of three-dimensional image structures to an object-centered, abstract description encoding shape features and structure relationships. We describe a prototype system that extracts three-dimensional (3-D) curvilinear structures from volume image data and transforms them into a symbolic description which represents topological and geometrical features of tree-like, filamentous objects. The initial segmentation is performed by 3-D line filtering and/or 3-D hysteresis thresholding. A center-line representation is derived by 3-D binary thinning and by compilation (raster-to-vector transformation) into a vector description. The final graph data-structure encodes the spatial course of line sections, the estimate of the local diameter, and the topology at important key locations like branchings and end-points. The analysis system is applied to the characterization of the cerebral vascular system segmented from magnetic resonance angiography (MRA).