Print

Print
Hi Bianca,

I don't think it makes a huge difference either way: ICs that are highly idiosyncratic tend not to appear in the group ICA. At the same time, patterns that are common for WM and CSF do tend to appear as components on their own.

I think I would regress to get rid of the "intermediate" cases, that could affect or interact with ICs of potential interest, but I don't think there is an obvious answer to this.

All the best,

Anderson


On 30 July 2017 at 06:04, Bianca De Blasi <[log in to unmask]> wrote:
Dear experts,

I am running a pipeline for the analysis of resting state fMRI data which includes FEAT+ICA_AROMA (as in the Manual), group ICA (temporal concatenation) and dual regression + randomise.

I am wondering whether it is recommended to carry out nuisance regression of WM and CSF before group ICA or if these components are automatically separated during ICA (and classified as different ICs) and hence will not affect the networks of interest.

Thank you,
Kind regards,
Bianca