help
help > RE: bypassing segmentation and use of masks
Apr 22, 2015 12:04 AM | Alfonso Nieto-Castanon - Boston University
RE: bypassing segmentation and use of masks
Hi Kajsa,
If you want to skip the segmentation and denoising steps I would suggest the following:
a) in Setup.rois, enter any arbitrary ROI file as Gray/White/CSF masks (e.g. conn/utils/otherrois/V1.img; any ROI will do, since these are going to be later disregarded) and run the Setup step normally
b) in Denoising, remove all of the effects in the 'Confounds' list, enter [0 inf] as the band-pass filter, select 'no detrending' and 'no despiking', and run the Denoising step
After that that you can continue to first-level analysis normally
That said, denoising, which removes noise, physiological, outlier, and subject-movement effects from the BOLD signal, is an integral part of obtaining valid connectivity measures, and without it connectivity measures are expected to contain relatively strong biases due to contribution of all of these effects, so in general I would strongly advise against skipping these steps unless you have already preprocessed your functional data to address all of these issues in some other manner.
Hope this helps
Alfonso
Originally posted by Kajsa Igelstrom:
If you want to skip the segmentation and denoising steps I would suggest the following:
a) in Setup.rois, enter any arbitrary ROI file as Gray/White/CSF masks (e.g. conn/utils/otherrois/V1.img; any ROI will do, since these are going to be later disregarded) and run the Setup step normally
b) in Denoising, remove all of the effects in the 'Confounds' list, enter [0 inf] as the band-pass filter, select 'no detrending' and 'no despiking', and run the Denoising step
After that that you can continue to first-level analysis normally
That said, denoising, which removes noise, physiological, outlier, and subject-movement effects from the BOLD signal, is an integral part of obtaining valid connectivity measures, and without it connectivity measures are expected to contain relatively strong biases due to contribution of all of these effects, so in general I would strongly advise against skipping these steps unless you have already preprocessed your functional data to address all of these issues in some other manner.
Hope this helps
Alfonso
Originally posted by Kajsa Igelstrom:
Hello,
I am wondering if it is possible to completely bypass the segmentation step without having to provide mask files for WM/GM/CSF. Basically, I need to skip all preprocessing and denoising. I have extracted seed time courses outside of CONN, and I want to run the analysis on a smaller part of the brain that has been isolated and aligned separately. The built-in segmentation can't cope with this partial brain coverage, and I can't (and don't want to) make segmentation masks for this part of the brain.
Before giving up I wanted to check with you in case I missed something obvious :)
Many thanks!
Kajsa
I am wondering if it is possible to completely bypass the segmentation step without having to provide mask files for WM/GM/CSF. Basically, I need to skip all preprocessing and denoising. I have extracted seed time courses outside of CONN, and I want to run the analysis on a smaller part of the brain that has been isolated and aligned separately. The built-in segmentation can't cope with this partial brain coverage, and I can't (and don't want to) make segmentation masks for this part of the brain.
Before giving up I wanted to check with you in case I missed something obvious :)
Many thanks!
Kajsa
Threaded View
| Title | Author | Date |
|---|---|---|
| KM Igelstrom | Apr 21, 2015 | |
| Alfonso Nieto-Castanon | Apr 22, 2015 | |
| Samantha Baldi | Oct 11, 2023 | |