Effects of motion and white matter hyperintensities on reference region based ¹⁸F-SynVesT-1 PET quantification in Alzheimer’s disease and healthy volunteers

Abstract

18 F-SynVesT-1 is a PET tracer targeting synaptic vesicle protein 2A, a surrogate marker of synaptic density. Simplified quantification using pseudo-reference regions such as the centrum semiovale (CSO) or whole cerebellum (CER) has previously been validated in healthy volunteers and Alzheimer’s disease. This study assessed how motion correction and white-matter hyperintensities (WMH) affect 18 F-SynVesT-1 quantification. We reconstructed 18 F-SynVesT-1 images for 40 healthy volunteers and 31 patients with Alzheimer’s disease 60-90 minutes post-injection with 3 motion handling methods: no motion correction; 5 minute frame-based post-reconstruction coregistration and averaging, and a high-temporal-resolution data-driven motion correction applied during reconstruction. WMH were segmented from T2-FLAIR images and excluded from the CSO to generate lesion-corrected CSO regions. A region-based voxelwise partial volume correction was applied. SUV and SUVR CSO/CER were compared across motion- and lesion-handling methods and the effects on group differentiation were explored. Data-driven motion correction during reconstruction improved visual image sharpness and increased SUV in high-binding cortical regions (median +0.3-1.0%, p≤0.001) while reducing SUV in the low-binding CSO (median -0.5%, p=0.001). Coregistering and averaging of frames only increased contrast when there was severe motion. The cerebellum was one of the least affected regions by motion correction (median SUV +0.4% with PET DDLM , p<0.001). WMH affected CSO SUV only when ≥25% of the region contained lesions and never affected the cerebellar SUV. SUVR CSO yielded better group differentiation than SUVR CER after accounting for motion and WMH. Individual patients’ z-scores differed significantly in both directions depending on the use of SUVR CSO or SUVR CER (e.g. mean absolute difference of 1.00 \(\pm\) 0.82 in the precuneus). Data-driven motion correction during reconstruction reduced the impact of head movements and attenuation mismatches, improving image contrast and quantitative accuracy. SUVR CER was less sensitive to motion and WMH than SUVR CSO , indicating greater robustness. When motion and WMH effects were controlled, SUVR CSO provided slightly better group differentiation than SUVR CER . These findings suggest that CSO may be preferred for group comparisons in well-controlled datasets, while the cerebellum represents a more robust reference region in the presence of motion or substantial WMH burden. The study was registered on ClinicalTrials.gov (NCT05384353), first posted on 20 May 2022.