DTI Smoothing by Hierarchical, Adaptive and Robust Strategy

Introduction Diffusion tensor imaging (DTI) has been widely used to construct the orientation and structure of fibers in biological tissues, particularly in the white matter of the brain [1]. The raw diffusion-weighted images (DWI) from which diffusion tensors are estimated, however, inherently contain large amounts of noise, leading to uncertainty in the estimation of the tensors and their derived quantities, including principal directions. Because many algorithms for fiber tracking reconstruct the directions of fiber pathways based on the principal directions of these estimated tensors, reducing the effects of noise on these estimated tensors is crucial for accurate tracking of fibers.