Nuisance effects in inter-scan functional connectivity estimates before and after nuisance regression

Abstract

In resting-state functional MRI, the correlation between blood-oxygenation-level-dependent (BOLD) signals across brain regions is used to estimate the functional connectivity (FC) of the brain. FC estimates are prone to the influence of nuisance factors including scanner-related artifacts and physiological modulations of the BOLD signal. Nuisance regression is widely performed to reduce the effect of nuisance factors on FC estimates on a per-scan basis. However, a dedicated analysis of nuisance effects on the variability of FC metrics across a collection of scans has been lacking. This work investigates the effects of nuisance factors on the variability of FC estimates across a collection of scans both before and after nuisance regression. Inter-scan variations in FC estimates are shown to be significantly correlated with the geometric norms of various nuisance terms, including head motion measurements, signals derived from white-matter and cerebrospinal regions, and the whole-brain global signal (GS) both before and after nuisance regression. In addition, it is shown that GS regression (GSR) can introduce GS norm-related fluctuations that are negatively correlated with inter-scan FC estimates. The empirical results are shown to be largely consistent with the predictions of a theoretical framework previously developed for the characterization of dynamic FC measures. This work shows that caution must be exercised when interpreting inter-scan FC measures across scans both before and after nuisance regression.

Resting-state connectivity stratifies premanifest Huntington’s disease by longitudinal cognitive decline rate

“…had been applied to the fMRI time series. However, no method is able to completely mitigate motion-related artefacts 44 , so we additionally controlled in stratification models for the effects of motion by removal of motion variability across subjects in each feature, as proposed by others 45 . The mean framewise displacement was used as a quantifier of head motion 44 , as provided by FSL's mcflirt, which calculates the average voxel displacement between successive time points in terms of rigid transformations over a spherical, brain-sized region of interest.…”