Hi,
I would like to use FSLNets to generate network matrices using the resting state networks identified in Group ICA (from MELODIC) as nodes in the matrix. I had a few questions, which I can't seem to find answers to elsewhere:
(1) FSLNets requires the output of Stage 1 of dual_regression as one of its inputs. Dual_regression requires a design.mat and design.con file as inputs. Given that dual regression first performs spatial regression using the Group ICs as inputs, does this mean that my design.mat file should be comprised of EVs that are the 1-D versions of the Group ICs? In my case, my group ICA has 70 ICs; should I input 70 z-maps as 70 EVs?
(2) If this is correct, given that the group MELODIC output generates a 4D .nii file containing the 70 ICs, do I need to split this 4D file into 70 3D files and then also transform these 3D files into 1D vectors to be used as inputs in the dual regression design matrix?
(3) Why should it be necessary to undertake the steps outlined above, given that dual_regression also takes the 4D group IC maps as an input? i.e. the 4D .nii file should contain all the info that dual_regression needs.
NOTE: I realize that there is an option to input "-1" in place of the design.mat and design.con files, however I noticed that the cross-subject stats with net mats requires the design.mat and design.con files as inputs, so I will have to generate these files anyway
(4) I have 4 different conditions (i.e. 4 different sleep stages/4 different groups) and I generated 4 separate Group ICA results. I'm not quite clear on the relationship between design.con and FSLNets. What I was hoping to do was generate 4 separate Netmats, one for each stage (though I could use the same Group RSN template (e.g. from Wake) for each NetMat); when generating a dual regression output, am I required to input the data for all 4 conditions into one implementation of dual_regression? I would expect to instead perform four separate implementations of dual_regression, but if that is the case, what goes into the design.con file? I don't see what is being contrasted...
(5) Why does FSLNets also require the output from slices_summary performed on the melodic_IC.nii Group IC maps? What does slices_summary provide FSLNets with, that isn't contained within the 4D group IC .nii file?
Thanks very much for your assistance!
|