Performance on all cognitive measures did not differ between individuals who had taken the tests once or three times (P > 0.10), nor did the number of visits interact with group (P > 0.42). Despite these negative results, number of visits was still included as a covariate in the remaining analyses to adjust for its potential minor influences Inhibitors,research,lifescience,medical on the cognitive measures. MRI acquisition and preprocessing All scans were obtained using a standard multimodal
protocol that included an axial 3D volumetric spoiled-gradient echo series (∼1 × 1 × 1.5 mm voxels) and a dual echo proton density/T2 (∼1 × 1 × 3 mm voxels) series. Thirty sites used General Electric 1.5 Tesla scanners, and two sites used Siemens 1.5 Tesla scanners. Each multimodal scan series was processed through BRAINS (Brain Research: Analysis of Image, Networks, and Systems) AutoWorkup (Pierson et al. 2011), a standardized morphometric processing pipeline Inhibitors,research,lifescience,medical that corrected for common multisite data differences (Magnotta et al. 2002). Outputs from the processing pipeline included basal ganglia volumes (caudate, putamen), a brain mask used for computing the intracranial volume (ICV), and a T1-weighted image, which was used in FreeSurfer for cortical Inhibitors,research,lifescience,medical thickness processing (Fischl et al. 2002). FreeSurfer estimates of cortical thickness demonstrate
very good test–retest reliability across scanners and sites (Han et al. Inhibitors,research,lifescience,medical 2006; Dickerson et al. 2008; Jovicich et al. 2009; Reuter et al. 2012). The brain mask was derived from all three image intensity modes to obtain robust estimates of ICV, which include tissue and surface cerebrospinal fluid (CSF) that extends to the border of dura mater. To account for individual differences
in head size, basal ganglia volumes were divided by ICV. The T1-weighted image Inhibitors,research,lifescience,medical was created with isotropic (1.0 mm3) voxels. T1 images were normalized so that the tissue intensities across the spatial domain of a single image and scans from different sites were placed in a consistent intensity range. Spatial intensity inhomogeneities were removed by applying a parametric correction (Styner et al. 2000) that used estimates of the tissue intensities based on tissue find more classes from the multimodal tissue classification (Harris et al. 1999). Each scan’s intensity range was placed on a consistent L-NAME HCl scale by linearly scaling to maximize the dynamic range inside the brain region. A reoriented, inhomogeneity, and intensity-corrected T1 scan for each subject was then clipped to the brain mask to be used as input for cortical parcellation. Cortical reconstruction was performed using the FreeSurfer image analysis suite (http://surfer.nmr.mgh.harvard.edu), which is an automated tissue classification and segmentation software that exhibits good test–retest reliability across scanner manufactures and field strengths (Han et al. 2006). Each subject’s MRI was initially analyzed in original space using the following analysis pipeline.