Pre-processed images were subjected to tensor decomposition to derive scalar diffusion indices, FA and MD (Le Bihan et al., 2001). This computation was performed in native coordinate space using Analysis of Functional NeuroImages' (Cox, 1996) diffusion plug-in routine, 3dDWItoDT (Cox and Glen, 2006). This algorithm uses a non-linear estimation method that guarantees that the diffusion tensor is positive definite, and provides estimates of FA and MD that are robust to increasing noise at high b-values (Jones and Basser, 2004; Skare et al., 2000). FA and MD were examined with whole-brain voxelwise analysis using Tract-Based Spatial Statistics (TBSS; (Smith et al., 2006)), which is optimized for multi-subject comparison and localized analysis of diffusion measurements, addressing the limitations of standard tools for voxelwise analysis where inaccurate registration and smoothing can lead to ambiguity in data interpretation.
