Volumes of the dorsolateral prefrontal cortex (DLPFC), inferior frontal gyrus (IFG) and the hippocampus in each hemisphere were derived from the T1-weighted scans and are reported in mm3. All volumetric analyses were performed blind to participant identity. The

cytoarchitectural justification and landmarks used for the frontal volumetric measures have been published in detail elsewhere (Cox et al., 2014). Briefly, the frontal lobe regions-of-interest (ROIs) were manually delineated on consecutive coronal slices at 1.3 mm thickness on AC-PC aligned T1-weighted volume scans by one of the authors (SRC). Key landmarks were identified on each slice and boundaries drawn by connecting those sulci with straight lines. learn more The DLPFC was ventrally limited by the inferior frontal sulcus, and medially by the crown of the most medio-superior gyrus. Both DLPFC and IFG were limited anteriorly by the frontal pole (a coronal plane at the most

anterior extent of the cingulate sulcus or paracingulate sulcus where present), and posteriorly by a coronal plane at the most anterior extent of the precentral gyrus. The IFG was limited dorsally by the inferior frontal sulcus, and ventrally by the lateral orbital sulcus in more anterior slices, or the circular sulcus of the insula in more posterior slices. As such, the Brodmann areas NVP-BGJ398 (BA) broadly represented were BA46/9 (DLPFC) and BAs 44, 45 and 47 (IFG). Intra-class correlation coefficients (agreement; Shrout & Fleiss, 1979) and Bland-Altman analysis (Bland & Altman, 1986) were conducted based on the absolute regional volumes of 20 hemispheres, parcellated 2 weeks apart by the same rater (SRC) for IFG (ICC = .96, Bland Altman mean = .93, 95% C.I. = −11.81 to 13.67), and DLPFC

(ICC = .99, Bland Altman mean = 1.19, 95% C.I. = −5.16 to 7.54). The hippocampus was initially segmented automatically and then each output was manually edited. Initial automatic segmentation was conducted using FSL FIRST, in which the T1-weighted volume was registered to an age-appropriate template (Farrell et al., 2009) and then to an optimised sub-cortical mask. Visual assessment and manual editing of the object masks was then conducted either by one of the authors (NAR) with an intra-rater correlation co-efficient of .98. Movement artefact in the anterior portion of two MRIs prevented prefrontal volumetric analysis, leaving 88 T1-weighted scans for the frontal sub-regions. Automated segmentation of the hippocampi failed in one case, leaving 89 participants with manually-edited hippocampal volumes. After pre-processing the diffusion MRI data to extract the brain, and remove bulk patient motion and eddy current induced artefacts, mean diffusivity (MD) and fractional anisotropy (FA) parametric maps were generated for every subject using tools freely available in FSL (FMRIB, Oxford, UK; http://www.fmrib.ox.ac.uk).