On the analysis of functional PET (fPET)-FDG: Baseline mischaracterization can introduce artifactual metabolic (de)activations.

Functional Positron Emission Tomography (fPET) with (bolus plus) constant infusion of [¹⁸F]-fluorodeoxyglucose (FDG), known as fPET-FDG, is a recently introduced technique in human neuroimaging, enabling the detection of dynamic glucose metabolism changes within a single scan. However, the statistical analysis of fPET-FDG data remains challenging because its signal and noise characteristics differ from both classic bolus-administration FDG PET and from functional Magnetic Resonance Imaging (fMRI), which together compose the primary sources of inspiration for analytical methods used by fPET-FDG researchers. In this study, we present an investigation of how inaccuracies in modeling baseline FDG uptake can introduce artifactual patterns to detrended time-activity curve (TAC) residuals, potentially introducing spurious (de)activations to general linear model (GLM) analyses. By combining simulations and empirical data from both constant infusion and bolus-plus-constant infusion protocols, we evaluate the effects of various baseline modeling methods, including polynomial detrending, regression against the global mean time-activity curve, and two analytical methods based on tissue compartment model kinetics. Our findings indicate that improper baseline removal can introduce statistically significant artifactual effects, although these effects characterized in this study (~2-8%) are generally smaller than those reported by previous literature employing robust sensory stimulation (~10-30%). We discuss potential strategies to mitigate this issue, including informed baseline modeling, optimized tracer administration protocols, and careful experimental design. These insights aim to enhance the reliability of fPET-FDG in capturing true metabolic dynamics in neuroimaging research.