Segmentation of diffusion tensor imagery

Segmentation paradigms in diffusion tensor imagery (DTI) are discussed in this paper. We present a technique for determining paths of anatomical connectivity from the tensor information obtained in magnetic resonance diffusion tensor measurements. The basic idea is to construct optimal curves in 3D space, where the optimality criteria is based on the eigenvalues and eigenvectors of the tensor. These curves are constructed via partial differential equations computed using multiple level-set functions. We also discuss our current efforts in clustering in DTI.