Poor functional BOLD brain masks

[dlevitas]

What happened?

I've noticed that with a subset of my dataset, the brain masks are poorly generated. I've attached two screenshots that demonstrate the issue. I believe this is tied to #2731 , which appears to be a lingering issue.

I can try to re-run fmriprep with new working directories and not re-use the freesurfer outputs, to see if that alleviates the issue. Additionally, I see that fmriprep version 22.0.0 is released, so I can try using that version as well. I'll report back once those are finished.

What command did you use?

                unset PYTHONPATH; singularity run \
		-B ${bids_root_dir}:${bids_root_dir} \
		-B /N/dcwan/projects/irf/templateflow:/opt/templateflow
		-B ${work_dir}:${work_dir} \
		-B ${output_dir}:${output_dir} \
		/N/dcwan/projects/irf/containers/fmriprep-${fmriprep_version}.simg 			\
		$bids_root_dir $output_dir													\
		participant																\
		--participant-label $s 													\
		--skip-bids-validation 													\
		--md-only-boilerplate													\
		--fs-license-file $FREESURFER_HOME/license.txt							\
		--fs-subjects-dir $output_dir/sourcedata/freesurfer/sub-${s}			\
		--output-spaces MNI152NLin6Asym:res-2									\
		--nthreads $cores														\
		--stop-on-first-crash 													\
		--resource-monitor 														\
		--low-mem 																\
		--mem_mb $reduced_mem 													\
		--use-plugin $bids_root_dir/fmriprep_plugin_${s}.yml 					\
		--use-aroma 															\
		--aroma-melodic-dimensionality 100 										\
		--dummy-scans $dummy_scans 												\
		--verbose 																\
		--bids-filter-file ${bids_root_dir}/code/bids_filter_file.json 			\
		--output-layout bids 													\
		-w $work_dir/sub-${s}"

What version of fMRIPrep are you running?

21.0.2

How are you running fMRIPrep?

Singularity

Is your data BIDS valid?

Yes

Are you reusing any previously computed results?

FreeSurfer, Work directory

Please copy and paste any relevant log output.

N/A

Additional information / screenshots

[dlevitas]

I re-ran my dataset with version 22.0.0, using clean freesurfer and working directories. Unfortunately the brain mask issue persists.

The SDC workflow also doesn't appear to have done a good job and the resulting image appears to be clipped, which might be causing downstream issues.

Another group at our institution uses --use-syn-sdcon their data (fmriprep version 21.0.2) since they don't collect field maps, and their results look fine. I'll therefore try using --use-syn-sdc instead of using my spin-echo field maps.

[dlevitas]

Using the --use-syn-sdc did not solve my issue, but I was able to determine what has caused my problem.

For context, I collected a SBRef for each functional BOLD image. Additionally, I collected both magnitude AND phase, as I'm hoping to use the phase data for a secondary analysis. For whatever reason, my Siemens Prisma scanner output a SBRef for the magnitude and the phase (thus 2 SBRef for each functional BOLD image). I thought I was excluding any phase images from being used by fmriprep by including that information in my --bids-filter-file bids_filter_file.json; however, this doesn't appear to be the case, as fmriprep was using the SBRef from the phase data (part-phase_sbref) as the reference for the BOLD mask. My solution was to simply remove the SBRef phase data from my BIDS folder.

I'll therefore close out this issue, but open a new one regarding the issue with the bids-filter-file