Susana Muoz Maniega1, James D. Bridson2,
Wei Jie Jensen Ang2, Paul A. Armitage1, Catherine
Murray3, Alan J. Gow3, Mark E. Bastin4, Ian
J. Deary3, Joanna M. Wardlaw1
1Clinical Neurosciences, University of
Edinburgh, Edinburgh, United Kingdom; 2Medicine and Veterinary
Medicine, University of Edinburgh, Edinburgh, United Kingdom; 3Psychology,
University of Edinburgh, Edinburgh, United Kingdom; 4Medical
Physics, University of Edinburgh, Edinburgh, United Kingdom
We
compare probabilistic neighbourhood tractography (PNT), an automatic tract
segmentation method, with a well accepted tractography method using manual
seed placement and multiple region-of-interest (ROI) constraints. Tracts were
segmented in the same data set using both methods and mean values of FA and
MD compared. Mean differences between PNT and ROI methods were ≤10%,
comparable with the reproducibility obtained when ROI are manually placed by
different operators. PNT segmentation showed a reasonable agreement with the
more conventional ROI tract segmentation method, with the advantage of
removing operator dependency.