While reviewing our fMRIPrep HTML outputs, we’ve noticed some irregularities in the BOLD mask (red outline) overlaid on the distortion-corrected images. Specifically, there appear to be “cut-out” regions or divots in the mask, where portions of the brain are excluded. The distortion correction itself appears to have run fine, it is just the red outline that is strange.
This issue is present across multiple participants (both younger and older adults) and appears in different slices, without a clear or consistent pattern.
We are unclear what this red bold mask is actually used for and are wondering whether this reflects a known issue with mask generation or distortion correction, and whether it might impact downstream analyses. We did confirm there is no ‘holes’ in the final preprocessed and optimally combined ME functional images, but perhaps this may affect confound estimation or other processing steps?
I’ve included example images below for reference.
Thanks in advance for your help!
Command used (and if a helper script was used, a link to the helper script or the command generated):
PASTE CODE HERE
Version:
fmriprep v23.1.3
Data formatted according to a validatable standard? Please provide the output of the validator:
Can you provide the code you used, as prompted by the Software Support post template?
Also, are you sure this is related to SDC? Do you get the same holes when not running SDC? And did you use dedicated fieldmaps or SYN SDC? Also does upgrading to the most recent fMRIPrep work?
It looks like this is just a problem with calculating the brain mask. Perhaps you could calculate a new mask from the BOLD. Whether or not you need to calculate a brain mask depends on what your analyses are. For example, if you are just interested in functional connectivity between two regions that are resolved in the mask, then those holes won’t impact stuff. I don’t know about multi-echo specific processing downstream impacts. In general I’d recommend having accurate brain masks - I bet recalculating the masks and/or upgrading fMRIPrep will help.
I think it is related to SDC since this is the only place the mask is shown and we don’t see it anywhere else; however, it’s possible it’s not. We were thinking of generating our own masks but were concerned that this bold-ref mask is automatically being used for other fmriprep steps that we are unaware of.
In any case, here is some more information on this incase it is relevant to other readers: we are using dedicated AP-PA field maps. For our analyses, we are running event-related PLS.
See the code used below for the SDC: module load fmriprep/23.1.3
