Dear experts,
I want to do seed-based connectivity in striatal regions using dual_regression, i.e. instead of IC maps to use striatal ROIs.
I have also run an melodic ICA with a number of RSNs some of which are noise.
My query is whether it is valid to use the noise RSNs in the seed-based analysis to regress their signal out in this way:
1) fslsplit the melodic_IC file to get the components that are noise.
2) fslmerge -t those components with the striatal seed ROI to create (e.g. a file called caudate_plusnoise.nii.gz)
3) run dual regression with this as input
My assumption is that the noise group ICs will be regressed out for the striatal seed input, as would happen for a normal IC dual_regression.
However, as before you recommended that I remove the demean option at 1st stage for the seed-based analysis, the outputs of regressing out the noise components would be different and or invalid?
Many thanks,
Elias
|