Tract-Based Spatial Statistics (TBSS) [85] was used for voxel-wise comparisons of FA and MD values between groups. TBSS attempts to correct for inaccuracies that complicate interpretation of other voxelwise methods (e.g., voxel-based morphometry), such as weaknesses in alignment, smoothing, and localization of brain tissue [85]. Diffusion measurement images (FA and MD) were derived from the 3dDWItoDT program of the Analysis of Functional NeuroImages package (AFNI) [23], which fits the diffusion weighted data to a tensor model using a non-linear least squares approach to compute the three orthogonal eigenvectors and corresponding eigenvalues necessary for calculating the scalar FA and MD indices based on standard formulas [68]. Additional processing procedures used Oxford Centre for Functional Magnetic Resonance Imaging of the Brain (FMRIB) Software Package tools (FSL) [86].