Meeting Banner
Abstract #0772

A Robust Spherical Deconvolution Method for the Analysis of Low SNR or Low Angular Resolution Diffusion Data

Jacques-Donald Tournier1, 2, Fernando Calamante, 23, Alan Connelly1, 2

1Advanced MRI development, The Florey Institute of Neuroscience and Mental Health, Melbourne, Victoria, Australia; 2Department of Medicine, University of Melbourne, Melbourne, Victoria, Australia; 3Advanced MRI development, Florey Institute of Neuroscience and Mental Health, Melbourne, Victoria, Australia

Analysis of low SNR, low b-value, or low angular resolution DWI data is difficult using HARDI methods such as spherical deconvolution. We propose to improve the robustness of spherical deconvolution to handle such data by including Rician correction and a constraint on the smoothness along fibres along with the commonly-used non-negativity constraint. We demonstrate this method on the types of data mentioned above, and show significant improvements in the quality of the reconstruction.

Keywords

according achieved actual allow already although amplitude amplitudes analyses angular anisotropic applies approximate approximately assessed bias bottom brain briefly channel clinical coarse coefficients coil commonplace computer constrained constraint control converts convolution coronal corrected correction crossing currently dataset datasets deconvolution definition dell development deviation diagonal diffusion directional distance distances distribution element enforce ensure equation evaluates evident expectation expense expressed extend extended feasible filtered form formed framework general give greatly head health historical improve improved improvements include included incorporate initial intensities introducing isotropic iteration iterations latter least limited linear loss matrix measured medicine model negative negativity next noise nonetheless note onto opposite ordinary orientation orientations original overall overcome particularly periphery perpendicular poor predicted problem problems processed processing projected projection proposed readily reasonable reduced refers relatively resolution robust robustly routinely sampled sampling significantly slices smooth smoothing smoothness solved spatial spherical squares starting suggests term terms tracking trans trio types uniformly updated vector whole workstation