help
help > RE: Scrubbing / Denoising Issue
May 12, 2021 12:05 AM | Nick Bray
RE: Scrubbing / Denoising Issue
For anyone interested, I figured this out:
1) I exported second level covariates (i.e., max motion, mean motion, etc.) to SPSS and identified extreme outliers via boxplots
2) I did the same thing with MRIQC's data for DVARS and FD
3) Cross-referenced the list to confirm the same participants were (more or less) listed in both
4) Excluded participants in the second level covariate tab via the same we do it for the second-level analysis (*facepalm*) using NaN
5) Reviewed QA-plots and saw a significant improvement (i.e., almost all >95%)
Cheers,
Nick
Originally posted by Nick Bray:
1) I exported second level covariates (i.e., max motion, mean motion, etc.) to SPSS and identified extreme outliers via boxplots
2) I did the same thing with MRIQC's data for DVARS and FD
3) Cross-referenced the list to confirm the same participants were (more or less) listed in both
4) Excluded participants in the second level covariate tab via the same we do it for the second-level analysis (*facepalm*) using NaN
5) Reviewed QA-plots and saw a significant improvement (i.e., almost all >95%)
Cheers,
Nick
Originally posted by Nick Bray:
Hi Everyone,
I am conducting my second project with the CONN toolbox. I am working with a cross-sectional group (n = 105; 1 MRI session) of older adults with MCI. For context, I am using version 20.b of CONN and data was imported after being pre-processed in fMRIPRep.
I have reached the Denoising stage. My scrubbing parameters are (I THINK) quite high (229p). I have also reviewed the QC-FC associations via the QA plots and see that Mean GS change, Mean Motion, and valid scans % match with NH DECREASE after the current denoising stagey. My questions are as follows:
1) How concerning is this DECREASE? As per Alfonso's handbook, > or = 95% is considered the "goal" (https://www.researchgate.net/publication...)
2) A select # of participants are quite troublesome (i.e., scrubbing over half the volumes). Is there a way to exclude/ignore a single participant at the denoising level to evaluate its impact on the overall QA plots
3) As a follow-up to #2, I recognize there is no gold-standard for denoising but was wondering if there is a "guideline" on how many volumes should be scrubbed before a participant is removed entirely. It was briefly discussed in this thread: https://www.nitrc.org/forum/forum.php?thread_id=6528&forum_id=1144
Thank you in advance.
Cheers,
Nick
I am conducting my second project with the CONN toolbox. I am working with a cross-sectional group (n = 105; 1 MRI session) of older adults with MCI. For context, I am using version 20.b of CONN and data was imported after being pre-processed in fMRIPRep.
I have reached the Denoising stage. My scrubbing parameters are (I THINK) quite high (229p). I have also reviewed the QC-FC associations via the QA plots and see that Mean GS change, Mean Motion, and valid scans % match with NH DECREASE after the current denoising stagey. My questions are as follows:
1) How concerning is this DECREASE? As per Alfonso's handbook, > or = 95% is considered the "goal" (https://www.researchgate.net/publication...)
2) A select # of participants are quite troublesome (i.e., scrubbing over half the volumes). Is there a way to exclude/ignore a single participant at the denoising level to evaluate its impact on the overall QA plots
3) As a follow-up to #2, I recognize there is no gold-standard for denoising but was wondering if there is a "guideline" on how many volumes should be scrubbed before a participant is removed entirely. It was briefly discussed in this thread: https://www.nitrc.org/forum/forum.php?thread_id=6528&forum_id=1144
Thank you in advance.
Cheers,
Nick
Threaded View
| Title | Author | Date |
|---|---|---|
| Nick Bray | May 5, 2021 | |
| Nick Bray | May 12, 2021 | |
| Alfonso Nieto-Castanon | May 13, 2021 | |