P2‐445: EVENT‐BASED MODELING OF THE TEMPORAL ORDERING OF REGIONAL β‐AMYLOID DEPOSITION IN THE BRAIN

Unraveling the spatio-temporal trajectory of β-amyloid brain deposition may improve early diagnosis and support future secondary prevention trials in Alzheimer's disease. This study applied event-based modeling (EBM) to determine the temporal ordering of amyloid accumulation in cognitively healthy elderly subjects. Cognitively healthy subjects (n=190) from the EMIF-AD PreclinAD Study were included and underwent dynamic PET scans (0-30min, 60min break, 90-110min), started at injection (185±10%MBq of 18Fflutemetamol). Amyloid status (positive or negative) was determined by the majority vote of three independent readers who visually assessed the images blinded to clinical information. Quantitative (regional) non-displaceable binding potentials (BPND) and standard uptake values (SUVR) were calculated from parametric images using cerebellar gray matter as a reference region. EBM was fitted to BPND and SUVR results, where each regional value corresponds to an event. The control group was composed of negative scans; the affected group of positive scans. Twenty six subjects (13%) were visually classified as positive using BPND images, and 28 (14%) using SUVR images. Figure 1 shows the positional variance diagrams for the ordering of amyloid abnormality in this population for each metric. BPND based EBM showed better agreement with recent studies of early amyloid deposition patterns, and greater certainty in event ordering and staging placement (Figure 2a). Earliest BPND regions included the parahippocampal and fusiform gyri; latest consisted of thalamus, hippocampus, cuneus and lingual gyrus. In contrast, SUVR displayed higher uncertainty, suboptimal staging (Figure 2b), and regional contradictory ordering, e.g. lingual gyrus as earliest positive region and fusiform gyrus and parahippocampal and fusiform gyri at later stages. A high proportion (60% with BPND and 54% with SUVR) of subjects were classified as stage 0, in line with the small number of positive scans and confirming the presence of pathological changes missed by global classifications of β-amyloid status. Positional variance diagrams of the uncertainty in the maximum likelihood event ordering estimated by taking MCMC (Markov chain Monte Carlo) samples using the EBM to fit (a) BPnd and (b) SUVR data. Proportion of subjects in each category (normal vs abnormal) at each EBM stage. Proportion of negative scans in dark blue and positive scans in green. Each EBM stage on the x-axis corresponds to the occurrence of a new regional transition event. Stage 0 corresponds to no events having occurred and stage 20 is when all events have occurred. Events are ordered by the maximum likelihood event sequence for the whole population as shown in Figure 1. EBM reconstructed the ordering of amyloid accumulation in a data-driven manner from a cross-sectional cohort without the use of pre-defined cut-offs for positivity. As expected, the resulting event order using BPND outperformed SUVR, showing greater agreement with recent staging studies and suggesting higher sensitivity to early pathological changes.