Hello,

I am working on a pipeline to go from fmriprep to CONN. Currently I want to set it up to do all the preprocessing and denoising outside of conn using afni's 3dTproject. We want to be able to orthogonalize all the confounds (aCompCor, realignment parameters, fd, bandpass, and spatial smoothing). I understand that CONN can do the denoising and bandpass filtering concurrently (or with orthogonality), but is there any way to also include the spatial smoothing in this step? 

I had two thoughts about how to go about this:

1) Load in fmriprep-processed data into CONN along with seeds. Then load in confound file(s) including those created for spatial smoothing and run all together in CONN's denoising step.

2) Denoise the data first and then load into CONN. But this also requires loading in the denoised seed timecourses, and I'm not exactly sure how to do that. Would I just load in denoised ROI files for each subject, rather than loading in a set of ROIs? 


I understand the argument to extract seed timecourses from the unsmoothed data to avoid "spillage," but does that create any sort of bias in comparing a timecourse derived from unsmoothed data to that of smoothed data? How does the relationship between seed radius and whole brain smoothing affect the results? 

Best,
- Harris