Abstract
In DT-MRI, diffusion-weighted multislice echoplanar images (EPIs)
are processed to represent water diffusion characteristics as a
diffusion tensor, reflecting the amount of diffusion in 3D. However
imaging quality is generally compromised by several factors including
the number of imaging slices, averages, diffusion sensitization
steps (b-values), voxel size, and gradient directions, resulting
in suboptimal SNR. In this study, we focus on improving imaging
quality and SNR by denoising and reducing systematic and random
errors through nonlinear anisotropic regularization. The raw EPIs
are directly regularized through a newly proposed nonlinear anisotropic
diffusion regularization method in 3D utilizing the gradient vector
flow fields and its performance is compared to conventional 2D and
vector-valued 2D anisotropic regularization methods. The effects
of these variants of anisotropic regularization are examined through
the maps of color-coded fractional anisotropy and tracked neural
fibers. The results show that DT-MRI regularization using the proposed
3D anisotropic diffusion significantly improves the quality of fiber
tracking and diffusion indices such as fractional anisotropy.
Users
Please
log in to take part in the discussion (add own reviews or comments).