A framework for modelling and clustering randomly structured white matter fibre tracts in diffusion tensor imaging

Reliable modelling and clustering of white matter (WM) fibre tracts are essential for studies using diffusion tensor imaging (DTI) tractography. This paper presents a novel scheme for modelling and clustering randomly structured WM fibre tracts reconstructed from DTI tractography. In this study, the mathematical representation of WM fibre tracts is formed by incorporating the diffusion orientation information and geometric characteristics of fibre tracts into the model. The quantitative measurements are achieved by calculating the pairwise affinity score between every two WM fibre tracts. This affinity score is sensitive to the shape, location and length of WM fibre tracts. A matching scheme is developed for finding piece-wise correspondences between two random WM fibre tracts. Real DTI datasets are used to assess the proposed approach. Experimental results show that this technique can effectively separate multiple fascicles, which do not have equal length and a common region of interest (ROI), into plausible bundles.