Segmentation of brain structures using PDE-driven surface growing

This paper describes an innovative technique for segmentation of the thalamus (a brain structure) from high resolution MRI brain images. Based on a powerful PDE-driven surface growing, this method of segmentation works really well with highly noisy and low-contrast datasets. The presented technique is essentially rooted in the PDE-based surface flow technique. It starts with an initial seed - a real 3D surface model. The growing velocity is computed automatically by the simulation of the PDE flow on the to-be-segmented dataset. Our experiments have demonstrated that it is a very efficient and general segmentation algorithm for many different brain structures.