help
help > RE: Denoised nifti DATA as input to GIFT toolbox
Mar 3, 2020 09:03 PM | Alfonso Nieto-Castanon - Boston University
RE: Denoised nifti DATA as input to GIFT toolbox
Dear Noelia
That is a good question, and yes, those denoised timeseries can be directly entered into GIFT for further analyses. It is true that you can actually remove the same linear trends and/or remove the average BOLD signal per voxel within GIFT, but I would still recommend doing that as part of CONN denoising strategy first (mainly because, as argued in Hallquist 2013, sequential denoising procedures risk reintroducing effects already removed by other/previous steps). In this case, if you do not detrend as part of CONN denoising, and then you detrend as part of GIFT denoising, the latter step may re-introduce some noise correlated with other known confounding terms (such as aCompCor terms). In contrast, including detrending as part of CONN denoising, even if you still later detrend again the results as part of GIFT denoising, will not risk this same effect (as the denoised timeseries will already be orthogonal to linear trends).
Best
Alfonso
Originally posted by Noelia Martinez Molina:
That is a good question, and yes, those denoised timeseries can be directly entered into GIFT for further analyses. It is true that you can actually remove the same linear trends and/or remove the average BOLD signal per voxel within GIFT, but I would still recommend doing that as part of CONN denoising strategy first (mainly because, as argued in Hallquist 2013, sequential denoising procedures risk reintroducing effects already removed by other/previous steps). In this case, if you do not detrend as part of CONN denoising, and then you detrend as part of GIFT denoising, the latter step may re-introduce some noise correlated with other known confounding terms (such as aCompCor terms). In contrast, including detrending as part of CONN denoising, even if you still later detrend again the results as part of GIFT denoising, will not risk this same effect (as the denoised timeseries will already be orthogonal to linear trends).
Best
Alfonso
Originally posted by Noelia Martinez Molina:
Dear Conn-experts,
I´m planning to use the denoised nifti output from CONN as input to the GIFT toolbox. I used the function conn_matc2nii.m to convert the matc files in the results/preprocessing folder to get the denoised nifti files.
Are these indeed the correct denoised nifti images that I can feed into GIFT to run a group ICA analysis?
My concern is that in the denoising step I used linear detrending and in the GIFT toolbox there is one preprocessing step to remove the mean per timepoint that cannot be deselected. How having linear detrending prior to this removal of the mean per timepoint will affect the results? Should I repeat the denoising without linear detrending?
Thank you in advance for your help.
Best wishes,
Noelia
I´m planning to use the denoised nifti output from CONN as input to the GIFT toolbox. I used the function conn_matc2nii.m to convert the matc files in the results/preprocessing folder to get the denoised nifti files.
Are these indeed the correct denoised nifti images that I can feed into GIFT to run a group ICA analysis?
My concern is that in the denoising step I used linear detrending and in the GIFT toolbox there is one preprocessing step to remove the mean per timepoint that cannot be deselected. How having linear detrending prior to this removal of the mean per timepoint will affect the results? Should I repeat the denoising without linear detrending?
Thank you in advance for your help.
Best wishes,
Noelia
Threaded View
| Title | Author | Date |
|---|---|---|
| Noelia Martinez Molina | Mar 3, 2020 | |
| Alfonso Nieto-Castanon | Mar 3, 2020 | |
| Noelia Martinez Molina | Mar 4, 2020 | |