**2.3 MRI data analysis**

Voxel-based morphometry was implemented by using the Statistical Parametric Mapping software (SPM2) (Wellcome Department of Imaging Neuroscience, London, England; www.fil.ion.ucl.ac.uk) (Friston et al., 1995). First, images were spatially normalized to the Montreal Neurological Institute (MNI) space with the standard T1-MRI template (Mazziotta et al., 1995) implemented in the SPM2 program, and re-sliced into a final voxel size of 1×1×1mm3 using tri-linear interpolation. The spatially normalized images were then segmented into three compartments: gray matter, white matter and cerebrospinal fluid. Furthermore, a Jacobian determinant was not introduced to modulate the resulting gray matter images so the voxel's values indicate the absolute density of the local gray matter. Finally, the segmented gray matter images from VoR with PTSD, VoR without PTSD, and HC were smoothed with a 12-mm full-width at half-maximum isotropic Gaussian kernel (Ashburner & Friston, 2000). The result of between-groups comparisons of gray matter images were performed in the general linear model.

Because we are particularly interested in exploring increases/decreases in GMD in VoR with PTSD compared to VoR without PTSD and HC, two-sample t-tests were performed in the VBM analysis in a voxel-by-voxel manner. Consistent with previous studies (Liberzon et al., 2007; Hou et al., 2007), the significance threshold was set to p < 0.005 corrected for multiple comparisons with a minimal cluster size of >50 voxels. The significant regions were superimposed onto SPM2's standard T1-weighted brain images. Based on previous research (Milad et al., 2006; Carrion et al., 2001; Richert et al., 2006; Hakamata et al., 2007; Bonne et al., 2003; Lanius et al., 2005; Molina et al., 2010; Bremner et al., 2006, 2008; Geuze et al., 2008b), we hypothesized that compared with HC, VoR with PTSD would show gray matter abnormalities in the prefrontal, temporal, parietal and limbic regions. We used the small volume correction (SVC) tool in the SPM2 package with the specific purpose of restricting comparisons to specific voxels located in these regions. This approach permits the implementation of hypothesis-driven analyses with corrections for the pre-specified ROIs rather than corrections for the whole brain.
