TBSS analyses involved the following steps: To achieve initial alignment, FA maps were registered to an averaged FA template (FMRIB-58) in MNI-152 standard space using an affine-only registration. This was followed by a non-linear transformation into 1 mm cubic voxel dimensions using Image Registration Toolkit (Rueckert et al., 1999). Data were examined for laterality, orientation, and cross-subject anatomical alignment. Next, transformed images were averaged across subjects to create a mean diffusion image (FA), from which a white matter skeleton was derived, representing tracts common to all subjects. Individual transformed FA images were then projected onto the skeleton. To minimize partial-volume effects and areas of high inter-subject variability, values were thresholded at FA >0.2. FA values from individuals' nearest relevant tract center were assigned to the skeleton via a perpendicular search for the maximum FA value within local skeleton structure. This process accounts for residual misalignments between subjects after the initial registration and minimizes systematic differences in tract location between groups of subjects. These data formed the basis of voxelwise statistical comparisons. MD data were processed using the same nonlinear transformation, skeleton, and skeleton-projection vectors derived from the FA analysis (Smith et al., 2007).
