Trouble at rest: how correlation patterns and group differences become distorted after global signal regression

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Brain Connectivity Analysis

Background:

• Resting-state functional magnetic resonance imaging (RS-FMRI) is a powerful tool for studying intrinsic brain functional connectivity.
• Analyzing RS-FMRI data presents challenges in distinguishing signal from noise, impacting the reliability of inferences.
• Global signal regression (GSReg) is a commonly used but controversial preprocessing step in RS-FMRI analysis.

Purpose of the Study:

• To critically evaluate the impact of global signal regression (GSReg) on resting-state functional magnetic resonance imaging (RS-FMRI) analyses.
• To demonstrate how GSReg can fundamentally alter interregional correlations and group comparisons.
• To argue against the continued use of GSReg in RS-FMRI studies due to its significant biases.

Main Methods:

• Discussion and re-evaluation of the effects of global signal regression (GSReg) on RS-FMRI data.
• Utilizing an illustrative model and derived equations to formalize objections to GSReg.
• Comparing the impact of GSReg on local and long-range correlations.

Main Results:

• Global signal regression (GSReg) significantly alters resting-state correlations, impacting patterns of functional connectivity.
• GSReg introduces biases that vary across brain regions, depending on the true underlying correlation structure.
• The method can artificially create or obscure group differences in functional connectivity.

Conclusions:

• Global signal regression (GSReg) should not be used in resting-state functional magnetic resonance imaging (RS-FMRI) analyses.
• GSReg's detrimental effects extend beyond negative bias, fundamentally compromising the interpretation of brain connectivity.
• Similar concerns apply to other denoising methods that aggregate signals across network regions without adequate signal separation.