First results of an automated model-based segmentation system for subcortical structures in human brain MRI data

We propose an automated model-based approach for segmenting subcortical structures, especially the basal ganglia, from 3D human brain MRI images. Our hybrid approach combines an automatic tissue classification scheme, a landmark detection algorithm and a discrete active surface model (ASM) to achieve its goal. A priori knowledge is incorporated into the different stages of the model and is used to retain shape and neighborhood information about the structures of interest throughout the segmentation process. The system is designed to require as little user interaction as possible. Sample results are given for the Brain Web simulated brain data set and for a real data set