Quantifying AD‐related brain amyloid with linearised progression models: Model‐based vs. data‐based

Background Brain amyloid-β (Aβ) is the pathological hallmark of Alzheimer’s disease (AD). In logistic models, Aβ accumulation a sigmoid function time-since-disease-onset (TSDO) (figure 1). Previous positron emission tomography (PET)-based models vary onset(t50) and duration(r) globally; capacity(K) baseline(NS) regionally (Whittington2018). We confirm existing approaches propose more powerful ICA-based approach to quantify severity estimate TSDO. Method used 1071 18F-florbetapir standard uptake value ratio (SUVR) images from ADNI-2 study (adni.loni.usc.edu/data-samples/data-types/pet). Images were mapped into MNI space. Averages extracted using Harvard-Oxford brain-atlas. Whole-brain tracer-specific parameters (Jack2013) obtained literature Of 16 regional (each 4 varied either or globally), optimal Bayesian information criterion was found with global t50 r, NS K 1) values r = 6.16y 4.10y. Linearised maps by regressing SUVR onto sigmoid. also estimated these as independent components, 2-component ICA on maps. Both outcomes weighting factors map. compared weights model in ADNI, effect size measured Hedges' g between cognitively normal (CN), subjective memory complaints (SMC), mild cognitive impairment (EMCI/MCI/LMCI) AD groups. 3 longitudinal visits (N 112) OASIS-3 (see www.oasis-brains.org) both methods, Centiloid (Klunk2015) 11C-PiB PET images. Result Maps capacity had spatial correlation 0.86 2); baseline 0.95. ADNI groups 2.25 for K, 2.42 (1.46 SUVR). OASIS-3, 1.24 1.46 (global 0.15, 0.4). Conclusion demonstrate that linear can be brain PET; yield larger sizes than method differentiating measuring changes visits.