Hello,
I am using[i] conn[/i] (v19.c) to analyze functional connectivity during n-back tasks in caffeine and caffeine withdrawal conditions compared to placebo. Since our resting ASL data have indicated a global change in perfusion in GM during caffeine intake and caffeine withdrawal, I would like to use the global perfusion (indexed by the mean value across GM CBF in every session) for global signal regression to control for the variance of perfusion, instead of, as suggested in the previous articles and the manual, using GM as confounder in the denoising step.
A little basic info about our study: 20 subjects in a 3-condition within subject design, i.e. each of the 20 subjects underwent caffeine, caffeine withdrawal, and placebo condition. Each condition has one n-back fMRI session consisting 0-back and 3-back. With 2 missing sessions, in total I have 58 scans.
Now my question is: When I did the "typical" functional analysis in SPM, I used the 58 x 1 matrix of global perfusion (single value for each session of each subject) for global scaling by ANCOVA. I want to do something similar in [i]conn[/i]. I prefer not to use GM as a confounder for denoising simply because, on top of the known controversy of GSR, with the necessity in our case to control the confounding effect from the changes of global perfusion, using GM CBF sounds more straightforward. I don't know if this makes sense?
(just a little detail: by looking at the before/after distributions of connectivity values, as well as the before/after GS in the denoising step, an inclusion of WM and CSF has done pretty well, I didn't see much gain by including GM)
If it can do, the subsequent issue is: How do I set up these single-value within-subject covariate? My intuition is to look at 1st-level covariate, however it requires a time-series information (don't know how to indicate a single value); I also tried to put in[i] setup/condition[/i], but it doesn't seem for a single value for each session; and and [i]2nd-level covariate[/i] is for between subject (allowing 20 x 1 matrix only).
Thank you in advance for any advice.
Best,
Shiuan
I am using[i] conn[/i] (v19.c) to analyze functional connectivity during n-back tasks in caffeine and caffeine withdrawal conditions compared to placebo. Since our resting ASL data have indicated a global change in perfusion in GM during caffeine intake and caffeine withdrawal, I would like to use the global perfusion (indexed by the mean value across GM CBF in every session) for global signal regression to control for the variance of perfusion, instead of, as suggested in the previous articles and the manual, using GM as confounder in the denoising step.
A little basic info about our study: 20 subjects in a 3-condition within subject design, i.e. each of the 20 subjects underwent caffeine, caffeine withdrawal, and placebo condition. Each condition has one n-back fMRI session consisting 0-back and 3-back. With 2 missing sessions, in total I have 58 scans.
Now my question is: When I did the "typical" functional analysis in SPM, I used the 58 x 1 matrix of global perfusion (single value for each session of each subject) for global scaling by ANCOVA. I want to do something similar in [i]conn[/i]. I prefer not to use GM as a confounder for denoising simply because, on top of the known controversy of GSR, with the necessity in our case to control the confounding effect from the changes of global perfusion, using GM CBF sounds more straightforward. I don't know if this makes sense?
(just a little detail: by looking at the before/after distributions of connectivity values, as well as the before/after GS in the denoising step, an inclusion of WM and CSF has done pretty well, I didn't see much gain by including GM)
If it can do, the subsequent issue is: How do I set up these single-value within-subject covariate? My intuition is to look at 1st-level covariate, however it requires a time-series information (don't know how to indicate a single value); I also tried to put in[i] setup/condition[/i], but it doesn't seem for a single value for each session; and and [i]2nd-level covariate[/i] is for between subject (allowing 20 x 1 matrix only).
Thank you in advance for any advice.
Best,
Shiuan