Surfaces

[danjgale]

Lately I've been extracting most of my fmri data from CIfTIs or GIfTIs, and have code lying around that does basically what niimasker does (without the reports, etc). I think this would make niimasker a standout tool and probably unique in this sense.

What would need to happen:

• Probably create an images.py module that has FunctionalImage (currently in niimasker.py) and then the subclasses for each image type (NiftiImage, GiftiImage, CiftiImage). These would inherit FunctionalImage and just have their own loading code and extraction methods (setting the regressors should be exactly the same).
• With the above point, it might be useful to then rename niimasker.py to extract.py, since it will no longer contain all of the code to run
• This should be possible without any external dependencies (i.e. just using nibabel and numpy). I have code to extract the mean timeseries for each ROI that mimics connectome workbench's wb_command -cifti-parcellate in addition to vertex-level extraction for an ROI mask.
• Would have to figure out how to visualize the roi_image on the functional surface, just like is done with the nifti volumes. But, I don't know how useful this is with surfaces? Currently the volume plotting is done with nilearn and not sure if their surface plotting would be similar

Finally, addding surface extraction would fly not reflect the name--niimasker. I've been toying with the idea and think that nixtract (NeuroImaging eXtract) might be a nice name as a replacement that is truer to it's expanded use.