help
help > RE: clarification reg' confounds
Sep 5, 2015 09:09 PM | Alfonso Nieto-Castanon - Boston University
RE: clarification reg' confounds
Hi Liron,
For CONN to include detrending as part of the Denoising procedure you simply need to have that option selected in the corresponding field in the Denoising tab (as your project seems to have). Similarly for CONN to remove motion regressors or outliers as part of the Denoising step it needs to have the 'realignment' and 'scrubbing' effects, respectively, listed as part of the 'confounds' list in the Denoising tab (which your project does not seem to have). So in your project, Denoising will only remove aCompCor components and linear trends, but it will not remove motion or outliers from the data.
Just to clarify, running ART Preprocessing step will create a new first-level covariate containing the outlier-scans info, and running realignment (in addition to actually realigning your data) will create a new first-level covariate containing the subject motion parameters. But this does not remove the outlier scans or the motion regressors from your BOLD signal just yet. That actually happens during the Denoising step (and only if the associated 'realignment' and 'scrubbing' effects are listed as part of your 'confounds' list, which typically are by default).
Last, as to why specifically you might not have those 'realignment' and 'scrubbing' effects listed in your 'confounds' list, as would happen typically by default, my guess is that you may have run your project originally without any of those first-level covariates (without running neither realignment nor ART as part of your preprocessing steps), and only later added those steps. The first time your run your Setup step and get to the Denoising tab, if you have not yet otherwise specified your desired choice of confounding effects to be used (e.g. through batch commands), CONN will go ahead and "auto-populate" that list with the typical/default settings (that will include White/CSF effects, as well as any first-level covariate and condition effects in your project). Once this list is filled CONN will not automatically modify this list in any case (even if you add new first-level covariates), so you need to manually modify that list if you wish to change/add/remove any effects that should be regressed-out during Denoising (e.g. if you add new first-level covariates for realignment or ART-based outliers in the Setup tab, you then need to add those corresponding effects to the 'confounds' list in the Denoising tab manually to let CONN know that you wish those effects to be included as part of denoising). Sorry if this is a bit confusing
Hope this helps and let me know if you would like me to further clarify any of the above
Best
Alfonso
Originally posted by L R:
For CONN to include detrending as part of the Denoising procedure you simply need to have that option selected in the corresponding field in the Denoising tab (as your project seems to have). Similarly for CONN to remove motion regressors or outliers as part of the Denoising step it needs to have the 'realignment' and 'scrubbing' effects, respectively, listed as part of the 'confounds' list in the Denoising tab (which your project does not seem to have). So in your project, Denoising will only remove aCompCor components and linear trends, but it will not remove motion or outliers from the data.
Just to clarify, running ART Preprocessing step will create a new first-level covariate containing the outlier-scans info, and running realignment (in addition to actually realigning your data) will create a new first-level covariate containing the subject motion parameters. But this does not remove the outlier scans or the motion regressors from your BOLD signal just yet. That actually happens during the Denoising step (and only if the associated 'realignment' and 'scrubbing' effects are listed as part of your 'confounds' list, which typically are by default).
Last, as to why specifically you might not have those 'realignment' and 'scrubbing' effects listed in your 'confounds' list, as would happen typically by default, my guess is that you may have run your project originally without any of those first-level covariates (without running neither realignment nor ART as part of your preprocessing steps), and only later added those steps. The first time your run your Setup step and get to the Denoising tab, if you have not yet otherwise specified your desired choice of confounding effects to be used (e.g. through batch commands), CONN will go ahead and "auto-populate" that list with the typical/default settings (that will include White/CSF effects, as well as any first-level covariate and condition effects in your project). Once this list is filled CONN will not automatically modify this list in any case (even if you add new first-level covariates), so you need to manually modify that list if you wish to change/add/remove any effects that should be regressed-out during Denoising (e.g. if you add new first-level covariates for realignment or ART-based outliers in the Setup tab, you then need to add those corresponding effects to the 'confounds' list in the Denoising tab manually to let CONN know that you wish those effects to be included as part of denoising). Sorry if this is a bit confusing
Hope this helps and let me know if you would like me to further clarify any of the above
Best
Alfonso
Originally posted by L R:
Thank you very much Alfonso for this helpful
answer!!!
A few follow-up questions: I though my model included detranding and motion correction, as I added them during the set up and denoising steps (see screenshot attached), but they do not appear in the list of confounds. Are they still regressed out by aCompCor? Are they regressed out otherwise?
Second, reg' the outlier detection (ART-based scrubbing)- I run this step (as it is included in the default preprocessing). Is there any additional step I need to do for actual outlier removal? If I understand correctly, this would remove single EPIs based on motion and other BOLD-related artifacts, how then do I see which EPIs were removed by this step? And lastly, If I run this step in preprocessing and didn't take any further action concerning outlier removal, was there still removal of outliers taking place?
Thanks so much!
Enjoy labor day (:
L.
A few follow-up questions: I though my model included detranding and motion correction, as I added them during the set up and denoising steps (see screenshot attached), but they do not appear in the list of confounds. Are they still regressed out by aCompCor? Are they regressed out otherwise?
Second, reg' the outlier detection (ART-based scrubbing)- I run this step (as it is included in the default preprocessing). Is there any additional step I need to do for actual outlier removal? If I understand correctly, this would remove single EPIs based on motion and other BOLD-related artifacts, how then do I see which EPIs were removed by this step? And lastly, If I run this step in preprocessing and didn't take any further action concerning outlier removal, was there still removal of outliers taking place?
Thanks so much!
Enjoy labor day (:
L.
Threaded View
| Title | Author | Date |
|---|---|---|
| L R | Sep 3, 2015 | |
| Alfonso Nieto-Castanon | Sep 3, 2015 | |
| L R | Sep 5, 2015 | |
| Alfonso Nieto-Castanon | Sep 5, 2015 | |
| L R | Sep 9, 2015 | |