AbstractBackground18F‐Florbetaben (FBB) is a PET radioligand approved for determining binary amyloid status. According to the product label, FBB images may be acquired anytime 45‐130 minutes post injection. Protocols vary in the acquisition of FBB with respect to use of Early (50‐70 minutes‐post‐injection) or Late (90‐110 minutes‐post‐injection) frames. Identifying differences between Early and Late acquisitions of FBB data could aid in image interpretation and inform efforts towards harmonization across cohorts. We evaluated inter‐rater agreement of visual reads (VR) and differences in SUVR thresholds to characterize Early vs. Late frames.MethodFifty‐three subjects were studied including 23 patients meeting clinical criteria for amnestic mild cognitive impairment or mild Alzheimer’s disease and 30 cognitively normal controls. FBB images were acquired 50‐110 min post‐injection in 5‐min frames. Images were corrected for subject motion and separated into 50‐70 and 90‐110 min time windows. Three certified readers blind to cognitive status and time window provided a binary VR for each of the 106 data sets. Global standardized uptake value ratios (SUVRs) were calculated using a cerebellar gray matter reference. An ROC‐derived SUVR cut‐off value was determined for each window using a preliminary cohort (n=23) and evaluated in a larger cohort (n = 53).ResultInter‐reader agreement was similar in 50‐70 min and 90‐110 min scans, (Fleiss Kappa= 0.597 and 0.598). SUVRs from 50‐70 min and 90‐110 min scans were highly correlated (r = 0.99, p<0.0001). SUVRs were greater in visually positive than in visually negative scans for both time windows (Fig. 1; p<0.0001). SUVRs were still increasing at 50‐70 min but became but plateaued at 90 min. ROC‐derived SUVR cut‐off values (Fig. 1; SUVR50‐70=1.342, SUVR90‐110=1.524) largely agreed with the majority consensus VR (49/53; 92.5%) across both time frames.ConclusionEarly and Late frame data were highly correlated, and both provided robust quantitative differences between visually positive and negative Ab groups. However, imaging at 90‐110 min provides more stable SUVRs and therefore is preferred for measuring longitudinal change in amyloid burden.