Global signal regression strengthens associations between behavior and resting-state fMRI
Stand-By Time
Monday, June 26, 2017: 12:45 PM - 2:45 PMSubmission No:
1806
Submission Type:
Abstract Submission
On Display:
Monday, June 26 & Tuesday, June 27
Authors:
Jingwei Li1, Ru Kong1, Nanbo Sun1, Avram Holmes2, Mert Sabuncu3, B.T. Thomas Yeo1
Institutions:
1National University of Singapore, Singapore, Singapore, 2Yale University, New Haven, United States, 3Massachusetts General Hospital, Charlestown, MA
First Author:
Introduction:
Regressing global (or whole brain) signal during resting-state fMRI preprocessing is controversial (Murphy2009; Saad2012). Global signal regression (GSR) can distort resting-state functional connectivity (RSFC; Saad2012). The GS is also not just noise, but might also contain neural information (Wong2013; Yeo2015). However, GS is also extremely effective at removing imaging artifacts, including motion (Power2015). Here we investigate whether global signal regression changes the association between RSFC and 25 behavioral/personality measures in the Brain Genomics Superstruct Project (GSP).
Methods:
We considered 873 GSP subjects (Yeo2011; Holmes2015) with both resting-state fMRI data and 25 behavioral/personality measures. We considered two identical preprocessing pipelines, whether GSR was performed. For both pipelines, preprocessing involves skipping the first 4 frames of each run, slice-time correction, motion correction, nuisance regression (ignoring censored frames), censoring (with interpolation), and bandpass filtering. The data were projected on fsaverage6 and smoothed by 6mm. For both pipelines, the nuisance regressors consisted of white matter, CSF, motion, and their temporal derivatives. One pipeline also includes GS (and its temporal derivatives) among the nuisance regressors. For each subject, a 400 x 400 RSFC matrix was computed based on a 400-area cortical parcellation (Schaefer, under revision).
The association between behavior and RSFC was accessed using a variance component method within a linear mixed-effects model (Sabuncu2016):
y = Xβ + f + e,
where y is an S x 1 vector consisting of the behavioral score of S subjects. X is an S x 6 nuisance matrix consisting a vector of all ones, sex, education, age, ethnicity and mean framewise displacement of each subject. f ~ N(0,σf2F) is an S x 1 random vector, where F is an S x S functional similarity matrix (FSM). The i-th row and j-column of FSM measures the RSFC similarity between the i-th and j-th subjects, defined as the correlation between the lower triangular entries of their 400 x 400 RSFC matrix. e is an S x 1 noise vector, where each entry is Gaussian distributed with 0 mean and variance of σe2.
Both σf2 and σe2 were estimated from the GSP subjects using restricted maximum likelihood (Sabuncu2016). We define the following "connectometricity" measure:
m = σf2/(σf2+σe2)
Connectometricity measures the variance in individual differences in behavior that could be explained by individual differences in RSFC.
The association between behavior and RSFC was accessed using a variance component method within a linear mixed-effects model (Sabuncu2016):
y = Xβ + f + e,
where y is an S x 1 vector consisting of the behavioral score of S subjects. X is an S x 6 nuisance matrix consisting a vector of all ones, sex, education, age, ethnicity and mean framewise displacement of each subject. f ~ N(0,σf2F) is an S x 1 random vector, where F is an S x S functional similarity matrix (FSM). The i-th row and j-column of FSM measures the RSFC similarity between the i-th and j-th subjects, defined as the correlation between the lower triangular entries of their 400 x 400 RSFC matrix. e is an S x 1 noise vector, where each entry is Gaussian distributed with 0 mean and variance of σe2.
Both σf2 and σe2 were estimated from the GSP subjects using restricted maximum likelihood (Sabuncu2016). We define the following "connectometricity" measure:
m = σf2/(σf2+σe2)
Connectometricity measures the variance in individual differences in behavior that could be explained by individual differences in RSFC.
Results:
Figs. 1 and 2 show the connectometricity estimates of 25 behavioral and personality measures with GSR (blue), without GSR (green) and difference between GSR and no GSR (brown). The connectometricity estimates were not high, but were consistent with the literature. For example, connectometricity (variance explained) for mental rotation and Shipley vocabulary scores were around 0.4, which was comparable with the prediction of fluid intelligence in the Human Connectome Project (R2 = 0.25; Finn2015). RSFC with GSR explained on average 3.4% more behavioral variance than RSFC without GSR. While the difference was small, the improvement was consistent (p = 2.8e-4).
Conclusions:
Global signal regression is controversial. We found that global signal regression strengthened the association between behavioral/personality measures and RSFC, compared with no global signal regression. In other words, after global signal regression, individual differences in behavior and personality could be better explained by individual differences in RSFC. One caveat was the relatively low morphometricity estimates for many behavioral and personality measures, although the variance explained were consistent with the literature (e.g., Human Connectome Project Megatrawl https://db.humanconnectome.org/megatrawl/).
Modeling and Analysis Methods:
fMRI Connectivity and Network Modeling
Motion Correction and Preprocessing 1
Task-Independent and Resting-State Analysis 2
Keywords:
Data analysis
FUNCTIONAL MRI
Other - fMRI preprocessing, global signal regression, whole brain signal regression, fluid intelligence, Big Five personality, inhibition, intersubject variability
1|2Indicates the priority used for review
Would you accept an oral presentation if your abstract is selected for an oral session?
I would be willing to discuss my abstract with members of the press should my abstract be marked newsworthy:
Please indicate below if your study was a "resting state" or "task-activation” study.
By submitting your proposal, you grant permission for the Organization for Human Brain Mapping (OHBM) to distribute the presentation in any format, including video, audio print and electronic text through OHBM OnDemand, social media channels or other electronic media and on the OHBM website.
Healthy subjects only or patients (note that patient studies may also involve healthy subjects):
Internal Review Board (IRB) or Animal Use and Care Committee (AUCC) Approval. Please indicate approval below. Please note: Failure to have IRB or AUCC approval, if applicable will lead to automatic rejection of abstract.
Please indicate which methods were used in your research:
For human MRI, what field strength scanner do you use?
Which processing packages did you use for your study?
Provide references in author date format
Holmes, A. J., et al. (2015), ‘Brain Genomics Superstruct Project initial data release with structural, functional, and behavioral measures’, Sci. Data 2:150031, doi: 10.1038/sdata.2015.31
Murphy, K., et al. (2009), ‘The impact of global signal regression on resting state correlations: are anti-correlated networks introduced?’, Neuroimage, vol. 44, no.3, pp. 893-905.
Power, J. D., et al. (2015), ‘Recent progress and outstanding issues in motion correction in resting state fMRI’, Neuroimage, vol. 105, pp. 536-551.
Saad, Z. S., et al. (2012), ‘Trouble at rest: how correlation patterns and group differences become distorted after global signal regression’, Brain connectivity, vol. 2, no.1, pp. 25-32.
Sabuncu, M. R., et al. (2016), ‘Morphometricity as a measure of the neuroanatomical signature of a trait’, Proceedings of the National Academy of Sciences, vol. 113, no. 39, pp. E5749-E5756.
Schaefer, A., et al., ‘Local-Global Parcellation of the Human Cerebral Cortex from Intrinsic Functional Connectivity MRI’, under revision.
Wong, C. W., et al. (2013), ‘The amplitude of the resting-state fMRI global signal is related to EEG vigilance measures’, Neuroimage, vol. 83, pp. 983-990.
Yeo, B. T., et al. (2011), ‘The organization of the human cerebral cortex estimated by intrinsic functional connectivity’, Journal of neurophysiology, vol. 106, no. 3, pp. 1125-1165.
Yeo, B. T., et al. (2015), ‘Functional connectivity during rested wakefulness predicts vulnerability to sleep deprivation’, NeuroImage, vol. 111, pp. 147-158.