RT Journal Article SR Electronic T1 A simultaneous method of MRI-based partial volume correction and image registration in [11C]PiB PET JF Journal of Nuclear Medicine JO J Nucl Med FD Society of Nuclear Medicine SP 1731 OP 1731 VO 59 IS supplement 1 A1 Keisuke Matsubara A1 Masanobu Ibaraki A1 Toshibumi Kinoshita YR 2018 UL http://jnm.snmjournals.org/content/59/supplement_1/1731.abstract AB 1731Objectives: Partial volume effect (PVE) due to limited spatial resolution of PET scanner is a critical issue in quantification of amyloid burden with amyloid PET study. Imprecise registration between PET and MRI, occasionally caused in conventional coregistration processes, can be another error factor in the quantification of amyloid with MR-driven anatomical templates. Several previous studies have demonstrated MR-based partial volume correction (MR-PVC) can be useful to avoid PVE in amyloid PET images. However, MR-PVC is very sensitive to the imprecise registration. We aimed to demonstrate applicability and usefulness of newly-developed simultaneous method for PVC and PET-MR image registration, PVC with optimized registration (PoR), in [11C]PiB PET studies. Methods: Brief procedures of PoR are the followings: 1) registration of PET images to individual MR images; 2) PVC of coregistered PET images with geometric transfer matrix (GTM) to generate PV-corrected PET images (GTM-PET); 3) smoothing for GTM-PET images with point spread function (PSF) to PET resolution; 4) re-registration of uncorrected PET images to smoothed GTM-PET; 5) repeat 1) - 4) until convergence; 6) PVC with region-based voxel-wise (RBV) to derive voxel-wise PV-corrected PET images. We applied PoR to [11C]PiB PET data for 92 subjects, downloaded from the ADNI database. As reference, conventional coregistration between PET and MR images and PVC of PET images coregistered to MR T1 images with RBV were also performed. We compared translation and rotation estimated with each registration method, regional PV-corrected standardized uptake volume ratio (SUVR; reference region: cerebellum) between conventional method and PoR. %CoV (=SD/mean) was also compared as index for intra-region variability of SUVR. Anatomical labels parcellated with FreeSurfer was merged to 23 volumes of interest for PVC and extracting regional values. Results: Registration with PoR converged until five iterations in the most of subjects. Significant differences of translations between conventional registration and PoR with five iterations were observed [mean (range) of translation at z-axis: 0.74 (0.00 - 2.73)]. In visual assessment, quality of registration between PET and MR images was improved by PoR. Significant differences of SUVR between conventional method and PoR were observed in all regions [%difference at maximum: 62.3% on sensory motor cortex]. %CoV significantly decreased by PoR in parietal, occipital medial temporal and sensory motor cortices, insula, hippocampus and subcortical white matter, relative to conventional method. Conclusion: PoR method can solve both the issue of PVE and imprecise registration simultaneously, and thus can be useful for accurate quantification of amyloid burden in [11C]PiB PET.