Dear all,

I have questions about data cleaning tools in the CONN toolbox.

I preprocessed my data (task-based fMRI) in SPM already, and I have a stripy artifact that I would like to see if aCompCor can remove.  I see here (fmri-methods/denoising-pipeline)

```
conn_module('preprocessing',...

'functionals', {'/data/func.nii'}, ...
'covariates', struct(...
'names', {{'realignment','scrubbing'}},...
'files', {{'/data/rp_func.txt', '/data/art_regression_outliers_func.mat'}}),...
'masks', struct(...
'White', {{'/data/c2anat.nii'}},...
'CSF', {{'/data/c3anat.nii'}}), ...
'steps', {'functional_regression', 'functional_bandpass'}, ...
'reg_names', {'realignment','scrubbing','White Matter','CSF'}, ...
'reg_dimensions',[inf, inf, 5, 5], ...
'reg_deriv', [1, 0, 0, 0], ...
'bp_filter', [0.008 inf] )

```

that segmented images Conn expects are the `c` prefixed white matter and CSF.  

Two questions about this: 

1) Aren't the `c`-prefixed segments in subject space?  I'm a little confused about why I would give CONN normalized functional images and segmented images in subject space.  This leads me to my second question:

2) *Can* I do the denoising step before I normalize?  I would like to for a couple of reasons.  Normalization didn't work incredibly well on my data.  Internal landmarks are well aligned but the front, back, and top of the brain all tend to be a little funky.  I don't have a ton of ideas about what might fix it (coregistration went well), but maybe getting some of the noise out of my functional data would help.  And even if it doesn't, it makes me nervous about using white matter and CSF to define signals of no interest if I'm less confident about the boundaries of those tissues in the normalized brain than I'd hope.  I guess another possibility is to create more conservative maps, but if it's permissible I'd like to perform this in subject space and keep normalization out of it.

I also have questions about ArtRepair to generate the `art_regression_outliers_func.mat` file. 

3) At what stage should I generate this file?  After slice time correction and realignment?  (I do slice time correction first.). Must my realigned files be resliced in order for it to work? Years ago I see from my own work that I entered smoothed realigned/slice time corrected (but not normalized) files.  I think someone familiar with CONN gave me that advice, but I can't find it reproduced anywhere on the internet.

4) Also, if I use this file in denoising, does it also need to be a first level regressor in SPM, or is this then redundant?

Thanks very much in advance for your help.

-Katie