NeuroMark PET: Replicable positron emission tomography ICA templates for florbetapir and florbetaben radioligands

Cyrus Eierud, Zening Fu, Helen Petropoulos, Anastasia Bohsali, Armin Iraji, Melanie Ganz, Cyril R. Pernet, Vince D. Calhoun

Abstract: Positron emission tomography (PET) tracer binding may not be aligned with commonly used parcellations of neocortex [1]. Independent component analysis (ICA) can capture coactivated regions among participants that might serve as robust templates from a data-driven perspective. NeuroMark is a framework combining pre-defined templates with spatially constrained ICA, capturing a wide range of brain markers across imaging modalities [2],[3]. Here, we build NeuroMark PET beta-amyloid templates for the 18F-florbetapir (FBP) and the 18F-florbetaben (FBB) radioligands, aiming to provide reliable spatial priors for future PET studies using ICA. Measuring the spatial similarities using correlation, the templates for the NeuroMark FBP and FBB templates matched strongly (ρ > 0.4) for 18 components. In a final step, the NeuroMark FBP template is applied on both FBB and FBP radioligands, showing that the age correlate profile across the FBB and FBP radioligands are statistically similar (p < 0.0008). The overall results demonstrate that NeuroMark ICA analyses are fully automated and the proposed PET templates would have great potential for large-scale analysis studies.