Dear FSL users/developers
I have run a group ICA with a dimensionality of 90. Several components at group level represent artefacts. I wonder if I should remove these components from melodic_IC.nii.gz using fslsplit and fslmerge before running dual regression or I should run dual regression with the full set of components and run randomise with only selected dr_stage2_ic files that are my RSNs of interest? In the literature I found examples for both options and I am not sure if one way is better than the other or equal.
Also, as a next step, I want to run FSLnets. Here, I can specify good (RSN) and bad (artefacts) components where bad components' time series would be regressed out from the good ones. Based on this it seems running dual regression on melodic IC file containing all components (artefacts and RSNs) is the preferable option.
Finally, I intend to regress out motion covariates and here I found two options, again. Should i use motion covariates as part of dual regression as done by Cole et al. 2012, 2013 or alternatively, I can use them as regressors in the statistical analysis (design matrix in randomise) as done by (Mowinckel et al., 2012, NeuroImage).
Thanks for your help
Best regards,
Judit Haasz
|