184 Chapter 7 uated inversion recovery (FLAIR) image (1.04 x 1.04 x 1.12 mm voxels, repetition time = 4800 ms, echo time = 278.8 ms, flip angle 90°). The 86 individuals from the baseline-only sample only underwent the above described dynamic [18F]flortaucipir PET and MRI scans at baseline. PET image analysis PET image analysis has been described previously [14, 31, 32]. Briefly, individual T1-weighted MRI scans were co-registered to native PET space, using Vinci software (Max Plank Institute, Cologne, Germany). The Hammers and Svarer templates incorporated in PVElab software were used to define cortical gray matter regions of interest (ROIs) on the co-registered MRI scans [33, 34]. Receptor parametric mapping (RPM) was applied to the PET data to obtain parametric images of BPND and R1, with the cerebellum gray matter as a reference region [35]. Previous research from our group has demonstrated that RPM is the best parametric method for [18F]flortaucipir [36], and has an excellent test-retest repeatability [9]. Tau PET data were additionally partial volume corrected using Van Cittert iterative deconvolution methods (IDM), combined with highly constrained back-projections (HYPR) as described previously [14]. Since results remained essentially unchanged when using partial-volume corrected (PVC) data, only non-partial volume corrected data are presented throughout the manuscript. We obtained BPND values (bilateral volume-weighted average) in three a priori defined regions corresponding to Braak staging regions of tau pathology (Braak I, Braak III/IV and Braak V/VI), as described previously [10]. For R1 we used all cortical gray matter ROIs as available in the Hammers template, with addition of the entorhinal ROI from the Svarer template. MR image analysis Cortical thickness (in mm) was obtained using FreeSurfer version 6.0.1 (https:// surfer.nmr.mgh.harvard.edu/). MR images from individuals with longitudinal data were processed through the recon-all longitudinal processing stream, including motion correction, skull-stripping, registration, segmentation, smoothing and parcellation mapping [37]. Images from individuals who only underwent baseline scans were processed through the recon-all processing stream for single timepoints. For both processing streams, cortical parcellation was performed based on the Desikan-Killiany Atlas (DKT), which contains 34 regions per hemisphere [38]. In order to improve cortical segmentation, a combination of T1-weighted + FLAIR images was used as input [39]. For individuals with missing FLAIR data at baseline and/ or follow-up, only T1-weighted images were used for cortical segmentation at all available time points (nAβ-= 1, nAβ+=3 out of follow-up sample (ntotal=61); nAβ+=5 out of baseline-only sample (ntotal=86)). Following all processing streams, segmentation
RkJQdWJsaXNoZXIy MjY0ODMw