Image Preprocessing and DTI Processing

For both cohorts, all DTI volumes were manually inspected to exclude the presence of large artifacts. FMRIB Software Library (FSL) (Smith et al., 2004) was used to perform all preprocessing steps and fit the DTI diffusion tensor model at each imaging voxel. Non-brain tissue was removed using FSL BET (brain extraction tool) (Smith, 2002), followed by motion and eddy-current distortions correction. Field maps were not acquired as part of these studies and thus corrections for susceptibility-induced distortions were not performed. For DTI analyses, FSL DTIFIT tool was used to compute diffusivities from fitting the diffusion tensor model and to generate DTI-FA (DTI-fractional anisotropy), DTI-MD (DTI-mean diffusivity), DTI-RD (DTI-radial diffusivity), and DTI-AD (DTI-axial diffusivity) volumes for each subject. DTI-derived image volumes for each participant were subsequently processed through the TBSS (Smith et al., 2006) pipeline to allow for whole-brain WM voxel-wise analyses as described below.

Since head motion during MRI scans is positively related to and shares genetic factors with BMI (Hodgson et al., 2017), and because registration-based correction methods do not exclude the effects of head motion entirely, we also computed motion parameters as described by Yendiki et al. (2014). These motion parameters include average volume-by-volume translation, average volume-by-volume rotation, percentage of slices with signal drop-out, and signal drop-out severity. In order to obtain these motion measures, we completed the image correction and quality assessment steps of the TRACULA pipeline (TRActs Constrained by UnderLying Anatomy), without running the WM pathways reconstruction steps (Yendiki et al., 2011). TRACULA-derived average volume-by-volume translation and average volume-by-volume rotation were included as regressors in subsequent voxel-wise and statistical analyses. The readout of percentage of slices with signal drop-out and signal drop-out severity were 0 and 1, respectively, for every participant in both cohorts.