First, six 5-min frames were identified and the last five of these frames were coregistered with the first, reducing effects of movement during the 30-min acquisition. These six coregistered frames were then averaged together and reoriented into a standard 160 × 160 × 96 voxel image grid with 1.5-mm cubic voxels. This image grid was oriented

such that the anterior–posterior axis of the subject was parallel to a line connecting Inhibitors,research,lifescience,medical the anterior and posterior commissures (the AC–PC line). Scans were then intensity normalized and smoothed with a scanner-specific filter function that was determined from phantom scans acquired during the certification process. This smoothing step corrected Inhibitors,research,lifescience,medical for differences between PET scanners and produced images with a uniform isotropic resolution of 8-mm full width at half maximum (FWHM). The downloaded scans were then spatially normalized to the SPM5 PET template (http://www.fil.ion.ucl.ac.uk/spm/). An average PET scan was generated from all

of the spatially normalized scans with Automated Image Registration (AIR, Woods et al. 1998). All further PET scan processing Inhibitors,research,lifescience,medical and analysis was performed using custom software written in MATLAB® (R2007b, The MathWorks, Natick, MA). The average PET scan was used to create a mask for extraction of brain voxels. The mask was defined as all SB431542 ic50 voxels with intensity >25,000. A single command in MATLAB® returns a vector containing all points at which a given comparison (e.g., >25,000) is true, ordered as if all

the columns in the volume were “unwound” into a single column. This vector of points can then be used as a list of indices for a new volume, Inhibitors,research,lifescience,medical thereby selecting only the points in the new volume that correspond to the points in the mask. All mathematical procedures were then undertaken on vectors created by selecting only the voxels within the mask. Statistical analyses were performed in R (R Development Core Team, Inhibitors,research,lifescience,medical 2008), using core routines and the lme4 module for linear mixed models. Mannose-binding protein-associated serine protease Significance testing for linear mixed models made use of Markov Chain Monte Carlo permutation analysis included in the languageR module. Projection and residual vectors In order to create a “query” vector for the identification of similarities between any given PET scan and those of patients with AD or MCI, it was necessary to isolate those aspects of AD PET scans that differ from normal PET scans. This distinction has traditionally been made using statistical comparisons of voxels or regions of interest (ROIs). One disadvantage of the traditional approach is that it is often necessary to perform numerous comparisons, which must be statistically corrected to avoid or minimize Type I errors.