Comparisons of fiber clustering algorithms for DTI images

Tractography is a promising technique to image brain white matter fiber tracts in diffusion tensor magnetic resonance imaging (DTI). However, the origin huge amounts of cluttered fibers are hard to be identified as different fiber structures with anatomical significance. A lot of fibers clustering methods have been proposed to automatically classify fibers, and in this paper, we focus on how to get an effective similarity measurement and efficient clustering algorithm for the fiber clustering. We introduce a framework for fiber clustering and results validation, and then evaluate the optimal combination method. Various combinations of the similarity measure and the clustering algorithm are implemented in the framework integrated with our visualization platform. Comparative experiments show that the best clustering performance and its corresponding similarity measurement and clustering algorithm.