Meeting Banner
Abstract #3555

Mask-Based Motion and Eddy-Current Correction of High B-Value Diffusion-Weighted Images

David Raffelt1, 2, J-Donald Tournier1, 3, Olivier Salvado2, Alan Connelly1, 3

1Brain Research Institute, Florey Neuroscience Institutes, Melbourne, VIC, Australia; 2The Australian E-Health Research Centre, CSIRO, Brisbane, QLD, Australia; 3Department of Medicine, University of Melbourne, Melbourne, VIC, Australia

Tractography and voxel-based analysis of diffusion-weighted images (DWI) benefit from higher-order models that require the acquisition of a large number of diffusion orientations at high b-values. The associated long scan times increase the likelihood of subject motion. Furthermore, high b-value DWI is also more prone to eddy-current induced distortions. Existing methods to correct motion and eddy-current artefacts are inadequate due to the poor SNR associated with high b-values. We present a robust method to correct motion and eddy-current artefacts in high b-value data. Validation experiments demonstrate that the proposed method can accurately correct b=3000s/mm 2 data with typical noise levels.

Keywords

accuracy acquisition added addition affine aligned allows alone amounts angular apply appreciably appreciated approaches assess assessed assuming atlas axis background blurry body brain clearly common commonly component components compose composite computed computing connected considerably contrasts convolution correct correcting correction corrects dataset datasets describe described differing diffusion dilation distance distortions durations eddy edge eliminates encode entirely erosion error evident experiment extent free furthermore generated gradient ground health holes illustrated illustrates improve improvement improvements in vivo induced influence institute institutes intensity inverting isotropic largest leading long longer marked mask masks metric models modulating motion mutual noise pair particularly performance populations prior process prone proposed pulse qualitative qualitatively registered registration relatively remaining reorient reoriented repeat repeats require resolve retaining rigid robust rotation scale sharper sharpness shear shears simulated software sonata spatial spherical squares steps subject subjects subsequently sufficiently suggests template thresholding towards transformation transformations translation trio truth typical update validate validation various volume volumes white yellow