Hongjian He1,2, Anna Leigh Rack-Gomer2,
Thomas T. Liu2
1Department of Physics, Zhejiang
University, Hangzhou, Zhejiang, China; 2Center for Functional MRI,
University of California, San Diego, La Jolla, CA, United States
Global
signal removal is a widely used and controversial method for resting state
functional connectivity analysis. When
all voxels are used for the computation of the global signal, removal of the
global signal can produce artifactual negative correlations. In this study, we consider the use of an
alternative estimate of the global signal that utilizes a random sample of
voxels chosen to be outside the regions of interest that are used to compute
the correlation. Because this estimate
does not include voxels within the regions of interest, its use does not
force negative correlations to exist.