An Adaptive Minimal Path Generation Technique for Vessel Tracking in CTA/CE-MRA Volume Images

We present an efficient method for the segmentation and axis extraction of vessels and other curvilinear structures in volumetric medical images. The image is treated as a graph from which the user selects seed points to be connected via 1-dimensional paths. A variant of Dijkstra’s algorithm both grows the segmenting surface from initial seeds and connects them with a minimal path computation. The technique is local and does not require examination or pre-processing of the entire volume. The surface propagation is controlled by iterative computation of border probabilities. As expanding regions meet, the statistics collected during propagation are passed to an active minimal-path generation module which links the associating points through the vessel tree. We provide a probabilistic basis for the volume search and path-finding speed functions and then apply the algorithm to phantom and real data sets. This work focuses on the contrast-enhanced magnetic resonance angiography (CE-MRA) and computed tomography angiography (CTA) domains, although the framework is adaptable for other purposes.