DTI imaging analysis

The Oxford Centre for Functional MRI of the Brain (FMRIB) Software Library (FSL 5.0, http://fsl.fmrib.ox.ac.uk/fsl/fslwiki/) [49–51] was used for raw DTI data analyses and calculations. First, for data pre-processing: (1) the EddyCorrect tool was used for correcting head motion and eddy current distortions by fine registration of the DTI images of the low-b (b value = 0 s/mm²) image; (2) brain masks of all the subjects were created using the Brain Extraction Tool (BET) [49]; and (3) using the least-squares algorithm fitting tensor model included in the DTI-FIT Tool [52], a diffusion tensor, or ellipsoid, was modeled at each voxel. Based on the eigenvalues of the tensor, FA, MD, AD, and RD values were calculated on a voxel by voxel basis.

Voxel-wise analysis of the DTI parameters (FA, MD, AxD, and RD) was performed using Tract-Based Spatial Statistics (TBSS) [53]. Using FSL's nonlinear image registration algorithm, all subjects' FA maps were aligned into a 1×1×1 mm³ standard Montreal Neurological Institute (MNI) 152 space. The target template was the FMRIB58_FA (http://www.fmrib.ox.ac.uk/fsl/data/FMRIB58_FA). Then a mean FA image was created by averaging the aligned FA maps. The mean FA image was thinned to create a mean FA skeleton representing the center of all tracts common to all participants in the present study. Each subject's aligned FA data were later projected onto the FA skeleton to obtain their FA skeletons and deformation matrixes. With the deformation matrixes, the skeletonized AxD, MD, and RD maps were created by the tbss_non_FA tool. The skeletonized FA, AxD, MD, and RD map images were subsequently fed to statistical analysis. ICBM-DTI-81 parcellation map [54] was applied to identify the names of WM tracts that contained the clusters of significant between-group differences. Index of ROIs from the ICBM-DTI-81 white-matter labels atlas followed by their abbreviations are present in Supplementary Table S8.