When I read the fsLR_den-91k_desc-denoised_bold.dtseries.nii file into MATLAB, I got a 91282 x T matrix in which the medial wall is already excluded. However, I still found that ~ 300 vertices had zero values across all time points for each subject. I am trying to understand the origin of these zero values and determine how best to handle them
Have you checked the fMRIPrep output that XCP-D ingested to see if those vertices are all zeros there?
@tsalo
Yes, it is. It turns out the issue has nothing to do with xcp-d. Perhaps I should ask a different question. Anyway, my initial thought was that the zeros might be caused by BOLD signal loss due to inhomogeneities in the B0 field, near OFC. However, after visualizing the data, I discovered that these zeros are located near the hippocampus.
When fMRIPrep does its surface projection, it excludes “bad” voxels if you use the --project-goodvoxels parameter. Here’s the relevant documentation for that parameter:
Exclude voxels whose timeseries have locally high coefficient of variation from surface resampling.
My guess is that you used that parameter and the goodvoxels mask flagged those voxels as bad, which is why the vertices are all zeroes in the CIFTI. It could be the result of something fMRIPrep does every without that parameter, but I don’t know enough about fMRIPrep’s surface projection to say.
You are right, I did include “–project-goodvoxels”. I further checked those zeros vertexs were located mostly in the cortex near hippocampus, brain stem and cerebellum. I guess I’ll just exclude those vertex. Thanks!
MANAGED BY INCF
