Gzip data corrupt error on files in work directory

1

Hi all,
I’m running fmriprep (version 20.2.5, using singularity) and getting the following errors on a couple of subjects:

indexed_gzip.indexed_gzip.CrcError: CRC/size validation failed - the GZIP data might be corrupt (file: /scratch/users/mayay/work/LMB_fmriprep/fmriprep_wf/single_subject_309_wf/func_preproc_ses_1_task_rest_run_1_wf/bold_std_trans_wf/_std_target_MNI152NLin2009cAsym.resnative/bold_to_std_transform/vol0343_xform-00343.nii.gz)

Node Name: fmriprep_wf.single_subject_309_wf.func_preproc_ses_1_task_rest_run_1_wf.bold_std_trans_wf.bold_to_std_transform

Looking at the html report, it seems that the pipeline ran smoothly until the AROMA step.
The error suggests that one of my nifti files is corrupted, but I’m confused since it doesn’t refer to my input files but for some intermediate derivative in the work folder, created by the fmriprep pipeline itself. Any ideas of where the problem might be? Happy to provide the log file and more details if anyone can help!

Thanks!
Maya

This is the command I’m running:
singularity run --cleanenv MYPATH/singularity_images/fmriprep-20.2.5.simg
/scratch/MYPATH/input
/scratch/MYPATH/derivatives participant --participant-label sub-309
-w /MYPATH/work/fmriprep
–output-spaces anat fsnative fsaverage MNI152NLin2009cAsym
–dummy-scans 4
–use-aroma \
–bids-filter-file /home/users/mayay/myScripts/filter_T1w.json

(the filter file selects a specific T1w image from the anat folder)

2

This is a bug in a dependency indexed_gzip. Upgrading to fMRIPrep 20.2.6 should resolve it. You should be able to reuse your working directory and pick up where you left off.

3

Thanks for answering so quickly!
Wouldn’t upgrading to the newer version lead to inconsistencies with my other subjects?

4

In general, upgrading within a minor series (20.2) should have minimal effect (see Bug-fix releases in the Nipreps docs). For fMRIPrep 20.2.5 to 20.2.6, the upgrade of indexed_gzip is the change that justified the release in the first place.

1 Like