Search

Aim: β-amyloid (Aβ) small animal PET facilitates quantification of fibrillar amyloidosis in Alzheimer's disease (AD) mouse models. Thus, the methodology is receiving growing interest as a monitoring tool in preclinical drug trials. In this regard, harmonization of data from different scanners at multiple sites would allow the establishment large collaborative cohorts and may facilitate efficacy comparison of different treatments. Therefore, we objected to determine the level of agreement of Aβ-PET quantification by a head-to-head comparison of three different state-of-the-art small animal PET scanners, which could help pave the way for future multicenter studies. Methods: Within a timeframe of 5 ± 2 weeks, transgenic APPPS1 (n = 9) and wild-type (WT) (n = 8) mice (age range: 13–16 months) were examined three times by Aβ-PET ([18F]florbetaben) using a Siemens Inveon DPET, a MedisonanoScan PET/MR, and a MedisonanoScan PET/CT with harmonized reconstruction protocols. Cortex-to-white-matter 30–60 min p.i. standardized uptake value ratios (SUVRCTX/WM) were calculated to compare binding differences, effect sizes (Cohen's d) and z-score values of APPPS1 relative to WT mice. Correlation coefficients (Pearson's r) were calculated for the agreement of individual SUVR between different scanners. Voxel-wise analysis was used to determine the agreement of spatial pathology patterns. For validation of PET imaging against the histological gold standard, individual SUVR values were subject to a correlation analysis with area occupancy of methoxy‑X04 staining. Results: All three small animal PET scanners yielded comparable group differences between APPPS1 and WT mice (∆PET=20.4 % ± 2.9 %, ∆PET/MR=18.4 % ± 4.5 %, ∆PET/CT=18.1 % ± 3.3 %). Voxel-wise analysis confirmed a high degree of congruency of the spatial pattern (Dice coefficient (DC)PETvs.PET/MR=83.0 %, DCPETvs.PET/CT=69.3 %, DCPET/MRvs.PET/CT=81.9 %). Differences in the group level variance of the three scanners resulted in divergent z-scores (zPET=11.5 ± 1.6; zPET/MR=5.3 ± 1.3; zPET/CT=3.4 ± 0.6) and effect sizes (dPET=8.5, dPET/MR=4.5, dPET/CT=4.1). However, correlations at the individual mouse level were still strong between scanners (rPETvs.PET/MR=0.96, rPETvs.PET/CT=0.91, rPET/MRvs.PET/CT=0.87; all p ≤ 0.0001). Methoxy-X04 staining exhibited a significant correlation across all three PET machines combined (r = 0.76, p < 0.0001) but also at individual level (PET: r = 0.81, p = 0.026; PET/MR: r = 0.89, p = 0.0074; PET/CT: r = 0.93, p = 0.0028). Conclusions: Our comparison of standardized small animal Aβ-PET acquired by three different scanners substantiates the possibility of moving towards a multicentric approach in preclinical AD research. The alignment of image acquisition and analysis methods achieved good overall comparability between data sets. Nevertheless, differences in variance of sensitivity and specificity of different scanners may limit data interpretation at the individual mouse level and deserves methodological optimization.

Highlights Microglial activation influences metabolic connectivity. Microglial depletion results in distinct decreases of metabolic connectivity. Metabolic connectivity increases in progranulin deficient mice. Cellular FDG allocation in progranulin deficient mice is shifted to microglia.