Meeting Banner
Abstract #3156

Multi-Fascicle Model Reconstruction from Acquisitions of DWI at a Single B-Value with a Population-Informed Prior

Maxime Taquet1, 2, Benoit Scherrer1, Benoit Macq2, Simon K. Warfield1

1Computational Radiology Laboratory, Harvard Medical School, Boston, MA, United States; 2ICTEAM Institute, Universit catholique de Louvain, Louvain-la-Neuve, Belgium

DTI are unable to distinguish fascicles crossing in one voxel, making them impractical to study the white matter microstructure and the brain connectivity. Multi-tensor models are a valuable option in these contexts. However, their reconstruction requires acquisitions of DWI at multiple b-values, unlike all datasets acquired with a single-shell HARDI sequence. This results in a waste of money, effort and imaging time. In this study, we propose a method to circumvent the multiple b-values requirement, allowing the community to use their existing datasets to reconstruct multi-fascicle models and conduct further analyses of the white matter at no additional cost.

Keywords

absence absolute according accuracy acquisition acquisitions adapted advantage affected align allows applying assess assisted atlas available bars best brain build built burden characterize circumvent close code comes common completed component computational computed constrained construct contaminates contrast corpus corrected correction crossing decrease decreased determined development diffusion distinguished eigenvalues eigenvectors errors estimation exact fascicle fascicles fiber financial fixed focus fraction fractions framework free full future gain gradient ground hence hypothesize identical improved incapacity infinite inflated informed integrate integration laboratory lowest macro medical model models motivated needs novel occluded ones optimize orientations others paired parameterized pathways performance poorly population potentially practical press preventing prior probabilistic problem proceed proceedings processing propose proposed radiology readily recent reconstruct reconstruction registration releasing replaced represent represented required response retrospectively school shell simplified since source spatially specifically squared step steps stick studies subject subjects subsets tensor tensors trace traces transform uncertainty uncorrected unlike uppermost validate vary volumes water white years