FMRIPREP: N4BiasFieldCorrection fails

[onu]

Dear fmripreppers,

I am experiencing weird EPI to T1 registrations for some FMRI sessions:

(this ‘was’ a whole brain scan of a healthy subject)

After going through the working directory, I think this comes from N4BiasFieldCorrection, because after this stage the output looks like the image above. Furthermore, calling N4BiasFieldCorrection with the fmriprep arguments and additionally –shrinkage-factor 3 seems to solve the problem.

This only happens for some of the MRI sessions, while others work fine.

I uploaded the command.txt and the input and output files for the N4BiasFieldCorrection call here.

I am happy to provide more information if needed,
onu

EDIT: I am using fmriprep version v1.3.1 with call
singularity run -B /data/pt_01994,/data/pt_01756/tmp/fmriprep_wd:/wd
/data/pt_01994/fmriprep-latest.simg
/data/pt_01994/bids
/data/pt_01994/fmriprep/output/
–fs-license-file /data/pt_01994/fmriprep/license.txt
participant --participant-label 01 02 03 04
–t2s-coreg
–output-space T1w template
–nthreads 10
-w /wda/
–fs-no-reconall


[oesteban]

Does the dataset have Single Band references? (sbref)

[onu]

Hello Oscar,

no we did not save sbrefs.
In the meantime I reran fmriprep for the subjects that had failed in the first go with –force-bbr and this ‘solved’ the problem.

[oesteban]

Glad you got that sorted out. Could you share the output log from fMRIPrep for the failing case? we would be interested in investigating the original cause of failure.

[onu]

There was no error detected, so there is no log directory in the fmriprep/sub-04 folder.
Or do you refer to the .html report?

I could ralso erun fmriprep with the –verbose flag if that helps.

[oesteban]

Yes, I meant the output from fMRIPrep. I think the argument -vvv will give us enough verbosity. Thanks!

[Alexander_Barnett]

I am experiencing the same issue as described above.

Re-running fmriprep with --force-bbr did not fix the N4 bias field correction. Outside of fmriprep, I called N4BiasFieldCorrection with the fmriprep arguments with the addition of --shrink-factor 3. That manages to produce a reasonable image.
Is it possible to run the N4BiasFieldCorrection inside the fmriprep call with the --shrink-factor 3 included?

[nickp]

Hi @oesteban,

We are experiencing a similar issue to @Alexander_Barnett, where in some subjects N4BiasFieldCorrection has failed, is there any update on a solution?

We have run reconall, BBR, and here is the output log for this sub:

[environment]
cpu_count = 4
exec_docker_version = “19.03.8”
exec_env = “fmriprep-docker”
free_mem = 11.6
overcommit_policy = “always”
overcommit_limit = “50%”
nipype_version = “1.5.0”
templateflow_version = “0.6.2”
version = “20.1.1”

[execution]
bids_dir = “/data”
bids_description_hash = “6ea419db713e2f79c762b87939f581089c1ed3aa5fb14688b2fe6ba629b5048d”
boilerplate_only = false
debug = false
fs_license_file = “/opt/freesurfer/license.txt”
fs_subjects_dir = “/out/freesurfer”
layout = “BIDS Layout: …/data | Subjects: 23 | Sessions: 46 | Runs: 0”
log_dir = “/out/fmriprep/logs”
log_level = 25
low_mem = false
md_only_boilerplate = false
notrack = false
output_dir = “/out”
output_spaces = “MNI152NLin2009cAsym:res-native”
reports_only = false
run_uuid = “20200806-075542_26cb7f18-f6e5-42ac-bda2-523910558694”
participant_label = [ “001”, “002”, “003”, “004”, “005”, “007”, “008”, “009”, “010”, “011”, “012”, “013”, “014”, “016”, “017”, “018”, “019”, “020”, “021”, “022”, “024”, “025”, “026”,]
templateflow_home = “/home/fmriprep/.cache/templateflow”
work_dir = “/scratch”
write_graph = false

[workflow]
anat_only = false
aroma_err_on_warn = false
aroma_melodic_dim = -200
bold2t1w_dof = 6
bold2t1w_init = “register”
cifti_output = false
fmap_bspline = false
force_syn = false
hires = true
ignore = []
longitudinal = false
medial_surface_nan = false
regressors_all_comps = false
regressors_dvars_th = 1.5
regressors_fd_th = 0.5
run_reconall = true
skull_strip_fixed_seed = false
skull_strip_template = “OASIS30ANTs”
skull_strip_t1w = “force”
spaces = “MNI152NLin2009cAsym:res-native”
use_aroma = false
use_syn_sdc = false

[nipype]
crashfile_format = “txt”
get_linked_libs = false
nprocs = 4
omp_nthreads = 3
plugin = “MultiProc”
resource_monitor = false
stop_on_first_crash = false

[seeds]
master = 56708
ants = 4767

[nipype.plugin_args]
maxtasksperchild = 1
raise_insufficient = false

Thanks!