Enhanced level-set approach to segmentation of 3-D heterogeneous lesions from dynamic contrast-enhanced MR images

A method for the 3-D segmentation of a heterogeneous (non-uniform intensity) volume of interest from high-resolution dynamic contrast-enhanced magnetic resonance image acquisition is presented and evaluated in this paper. The algorithm uses a level-set approach with enhanced edge sensitivity. The local probability near the lesion boundary is utilized for level-set evolution. The level-set function is sensitized to the boundary of the lesion by combining forces derived from the image, including gradient vector flow. The proposed algorithm performs better than other applicable segmentation techniques based on active contours, and also the classical level-set approach. Results are presented and analyzed to validate the same.