Meeting Banner
Abstract #1426

Hierarchical Diffusion Tensor Image Registration Based on Tensor Regional Distributions

Pew Thian Yap1, Hongtu Zhu2, Weili Lin1, Dinggang Shen1

1Department of Radiology and BRIC, University of North Carolina, Chapel Hill, NC, USA; 2Department of Biostatistics and BRIC, University of North Carolina, Chapel Hill, NC, USA

We propose a novel DTI registration algorithm which leverages on the hierarchical guidance of tensor regional distributions and local boundaries, both extracted directly from the tensors. This is in contrast with conventional methods which typically compute regional and edge information based on tensor scalar maps, which might not necessarily reflect the actual tensor information. For each voxel, statistical measures such as mean and variance are computed in various neighborhood sizes, extracting multiscale information. Edge boundaries are obtained by an extended Canny edge detector which works directly on the tensors. Distinctive features are then selected hierarchically as landmark points to guide the registration.