*2.3. Image Processing*

Images were processed using a combination of open-source software packages: MRtrix [36], FSL [37], and greedy [38]. Standard MR images (T1w, T1ce, T2w, and FLAIR) were non-linearly registered to the MNI152 (Montreal Neurological Institute, MNI) atlas [39] space using the deformable registration package greedy. Automated tissue-type segmentation was performed on T1w image sets using FSL-FAST (FMRIB's Automated Segmentation Tool) [40]. DT images were preprocessed to remove noise and corrected for distortion and field-bias using MRtrix's dwidenoise [41], dwipreproc [42], and dwibiascorrect [40,43] scripts. After preprocessing, DT images were linearly registered into T1w-space using FSL-FLIRT (FMRIB's Linear Image Registration Tool) [44] and then transformed into MNI-space by applying the affine matrix generated to register the T1w image. Lastly, all image sets were downsampled by a factor of 0.45 with cubic interpolation using MRtrix to avoid oversampling (voxel size: 1.75 mm3).
