An automated pipeline for Centiloid quantification of amyloid‐β using multiple 11C‐PiB‐PET and 18F‐PET tracers

Abstract

AbstractBackgroundQuantitative measures of amyloid‐ß (Aß) pathology using positron emission tomography (PET) imaging are sensitive to identify pathological changes early in Alzheimer’s disease (AD). The Centiloid scale aims to standardize these in vivo amyloid quantifications to a 100‐point scale, where an average value of zero signifies high certainty of amyloid negativity and 100 identifies average typical AD Aß‐pathology load (Klunk et al., 2015). The current study developed and validated a single and fully automated Centiloid quantification pipeline for multiple amyloid PET compounds.MethodQyScore’s® fully automated pipeline was validated on 11C‐PiB‐PET and 18F‐PET images from the Centiloid project (https://www.gaain.org/centiloid‐project): 34 young controls [age = 31.5 ± 6.3 years] and 47 AD patients [age = 67.5 ± 10.5 years; CDR = 0.5–1]. 18F tracers included Florbetapir (FBP, N = 46), Forbetaben (FBB, N = 35), Flutemetamol (FTM, N = 74) and NAV4694 (NAV, N = 55). PET/MR image pairs were both co‐registered and normalized in the MNI template space. The fully automated segmentation from QyScore®, a CE‐marked and FDA‐cleared neuroimaging medical device, parcellated the masks of the grey matter tissue (target) and of the cerebellum (reference) region (Figure 1). The standardized uptake value ratio (SUVr) was computed as the ratio of the mean signal in both regions. Correlations of 11C‐PiB‐PET and 18F‐PET SUVr values with published SUVr data were computed. Further, correlations between 18F‐PET SUVr and paired 11C‐PiB‐PET SUVr were computed. Correlation coefficients (R2) > 0.7 were required to consider the Centiloid calibration valid.ResultQyScore’s® fully automated quantitative pipeline produced SUVr values well within the bounds defined by the Centiloid method (SUVr_AD‐100 = 2.08 +/‐ 0.2; SUVr_YC‐0 = 1.01+/‐ 0.05; R2 = 0.99; slope = 1.00; intercept = ‐0.44). 11C‐PiB SUVr correlation coefficients with published values were above 0.99 (Figure 2). Correlation coefficients of 18F‐PET SUVr with 11C‐PiB‐PET SUVr were respectively 0.91, 0.95, 0.96, 0.99 (Figure 3). Equations for converting to Centiloid were respectively: • Centiloid = 177.79 FBP_SUVr ‐ 183.56 • Centiloid = 153.08 FBB_SUVr – 152.93 • Centiloid = 122.39 FTM_SUVr – 120.97 • Centiloid = 90.20 NAV_SUVr – 91.61ConclusionWe demonstrate the feasibility and reliability of a fully automated amyloid PET pipeline for multiple amyloid‐PET compounds (11C‐PiB and 18F) suitable for implementation in clinical trials.