Jakub Piatkowski1, Amos J. Storkey2,
Mark E. Bastin3
1Neuroinformatics Doctoral Training
Centre, University of Edinburgh, Edinburgh, United Kingdom; 2Institute
for Adaptive and Neural Computation, School of Informatics, University of
Edinburgh, Edinburgh, United Kingdom; 3Medical Physics, University
of Edinburgh, Edinburgh, Midlothian, United Kingdom
We
use a fully physical two-compartment model, comprising isotropic and
anisotropic terms, to describe diffusion MRI data. The posterior
distributions over the parameters of this model are estimated using sampling
techniques. This yields maps of white matter (WM) volume, which reveal a
level of structure missing in FA maps. Additionally, we get tensor parameters
for the anisotropic compartment (i.e. WM), which provide a measure of
fibre-specific anisotropy that doesn't suffer from partial volume effects.