Dear Dr. Alfonso and the CONN community,
I am currently analyzing resting-state fMRI data for 12 PPMI subjects using the default preprocessing pipeline in the latest version of CONN.
While reviewing the Denoising tab, I noticed that the functional brain mask (BOLD %) does not align correctly with the anatomical image for 2 subjects. One of them is a subject already excluded due to severe motion, which I can understand. However, the other one (Subject 3) was not excluded, yet still shows a severe misalignment (please see the attached image).
I found some forum posts mentioning this might be a centering or normalization issue. However, I ran the default preprocessing pipeline, which I believe already includes automatic centering. Furthermore, when I checked the 2nd-level QA covariates related to normalization, Subject 3 did not seem to show any obvious outlier flags.
Could you please advise why this misalignment occurred despite using the default pipeline? Is there a specific QA metric I am missing that should have caught this, and how can I manually fix this specific subject?
I have attached a single screenshot showing the misaligned brain map for your reference.
Thank you
-minsu kim-
Hi,
I ran into the same issue a few days ago (see: https://www.nitrc.org/forum/forum.php?thread_id=15985&forum_id=1144 )
I’m still having trouble fully understanding it and haven’t found a solution yet. In the meantime, could you let me know whether you were able to complete the analysis? If so, did you try overlaying the denoised functional image (dswau*.nii) onto the anatomical image (wc0c*.nii)? In my case, they never align (see Figure 1). Interestingly, the smoothed but non-denoised images (swau*.nii) consistently align perfectly with the anatomical image (see Figure 2).
FIGURE 1:
FIGURE 2:
Kind regards,
Adrian.
Dear Minsu Kim
Not sure what could have caused this, but it appears your anatomical image may have not gone through the standard preprocessing pipeline (the anatomical image in your attached figure should be already skull-stripped). Perhaps normalization/segmentation failed badly, but you are right that if that was the case that would have resulted in a notably lower value for that subject's QC_NORM_struct. Just for sanity, I would check perhaps in Setup.structurals whether that subject's anatomy is actually pointing to the fully-preprocessed anatomical images (e.g. named wc0cT1.nii) instead of pointing to the original images (e.g. named T1.nii) or some other intermediate files.
Hope this helps
Alfonso
Originally posted by mskim_usc:
Dear Dr. Alfonso and the CONN community,
I am currently analyzing resting-state fMRI data for 12 PPMI subjects using the default preprocessing pipeline in the latest version of CONN.
While reviewing the Denoising tab, I noticed that the functional brain mask (BOLD %) does not align correctly with the anatomical image for 2 subjects. One of them is a subject already excluded due to severe motion, which I can understand. However, the other one (Subject 3) was not excluded, yet still shows a severe misalignment (please see the attached image).
I found some forum posts mentioning this might be a centering or normalization issue. However, I ran the default preprocessing pipeline, which I believe already includes automatic centering. Furthermore, when I checked the 2nd-level QA covariates related to normalization, Subject 3 did not seem to show any obvious outlier flags.
Could you please advise why this misalignment occurred despite using the default pipeline? Is there a specific QA metric I am missing that should have caught this, and how can I manually fix this specific subject?
I have attached a single screenshot showing the misaligned brain map for your reference.
Thank you
-minsu kim-