Global signal regression strengthens association between resting-state functional connectivity and behavior

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Data Science

Background:

• Global signal regression (GSR) is a controversial preprocessing technique in resting-state functional MRI (rs-fMRI).
• While GSR removes artifacts, its impact on neural information and functional connectivity is debated.
• Previous research focused on artifact removal and potential biases of GSR.

Purpose of the Study:

• To investigate whether GSR strengthens or weakens the association between resting-state functional connectivity (RSFC) and behavioral measures.
• To assess the utilitarian impact of GSR on functional connectivity fingerprinting across cognition, personality, and emotion.

Main Methods:

• Applied variance component models and kernel regression to rs-fMRI data from the Brain Genomics Superstruct Project (GSP) and Human Connectome Project (HCP).
• Analyzed whole-brain RSFC associations with 23 behavioral measures in GSP and 58 in HCP.
• Evaluated the impact of GSR on behavioral variance explained and prediction accuracy.

Main Results:

• GSR increased the variance explained by whole-brain RSFC for behavioral measures by an average of 47% (GSP) and 40% (HCP).
• GSR improved behavioral prediction accuracies by 64% (GSP) and 12% (HCP).
• Improvements were consistent across different analytical methods and particularly benefited task performance measures.

Conclusions:

• GSR generally strengthens the associations between whole-brain RSFC and most behavioral measures in young, healthy adults.
• The observed improvements are unlikely due to motion or respiratory artifacts, as GSR effectively removes these.
• Further research is needed to determine GSR's impact on specific brain connections and networks.