Global signal regression has complex effects on regional homogeneity of resting state fMRI signal.

Regional homogeneity (ReHo) quantifies spatially local synchronization of resting state fMRI signal and has been applied to lots of clinic studies. Accumulating evidences demonstrated that the synchronization between spatially distinct brain regions, i.e. functional connectivity, can be remarkably influenced if the global mean time course is regressed out, namely global signal regression (GSR). Very recently, it was reported GSR reduces the test-retest reliability of ReHo, and reduces the positive correlation between ReHo and head motion. In this study, we were interested in two questions: 1) how GSR affects the raw ReHo values and its spatial distribution over the brain; 2) how GSR affects the differences of ReHo between two resting states, eyes open (EO) and eyes closed (EC), in healthy individuals. We found that the ReHo values were reduced by GSR but the spatial distribution of ReHo was not changed remarkably. In addition, split-half reproducibility analysis showed reproducible ReHo difference between EO and EC in some areas (e.g., thalamus and caudate) only with GSR, but showed reproducible ReHo difference in some other area (right temporal pole) only without GSR. The effects of GSR were almost independent of regression of other nuisance covariates. Our results suggest that the influences of GSR on ReHo are remarkable, reliable and complex. For the between-condition comparison, the GSR effects are region specific. We suggest that, for application studies using ReHo approach, it would be helpful to report results both with and without GSR.