I would not run ICA a that high of resolution on the surface, as it will
make it take a lot longer than needed. Instead, I would use a surface
mesh resolution more similar to that of the original data (for 2mm HCP
data we use 32k). ICA does not care about spatial connectedness, so one
can repackage the surface or CIFTI data into a NIFTI-1 file and run
melodic on it (autodimensionality estimation is not available however). I
think a beta version of melodic that runs natively with CIFTI will be
available in the next FSL release.
Until that point, you can use wb_command -cifti-convert -to-nfiti
-from-nifti to use the current version of melodic.
Peace,
Matt.
On 1/26/17, 11:53 AM, "FSL - FMRIB's Software Library on behalf of David
Parker" <[log in to unmask] on behalf of [log in to unmask]> wrote:
>Hello everyone!
> I'm running a surface analysis on my data using FsFast, and I would
>like to preform ICA on the vectorized surface timeseries data created by
>this process (using fsaverage as a template for data with N timepoints, a
>163842 x 1 x 1 N nifti file is created for the left and right hemisphere).
>
>Can I use melodic on these to extract ICAs from this data? I understand
>that it does dimensional adjustments anyways, but I'm not sure if it's
>designed to work on data like this. Any help would be appreciated,
>thanks!
|