fMRIPrep atropos(?) Error

ebaek · November 30, 2019, 10:03pm

Hello, I have a dataset that I’ve run through fMRIPrep where about 20 of the subjects (out of 126) crashed with similar errors.

I have tried to re-run the subjects with increased memory allocation, but it continues to crash with the same error. Any help would be greatly appreciated. Thank you!

The error at the end of the HTML output file of one of the subjects:

Node Name: fmriprep_wf.single_subject_018PJS_wf.anat_preproc_wf.brain_extraction_wf.atropos_wf.01_atropos

The error in the log:

Node: fmriprep_wf.single_subject_018PJS_wf.anat_preproc_wf.brain_extraction_wf.atropos_wf.01_atropos
Working directory: /u/project/cparkins/data/kshap_cyberball/work/fmriprep_wf/single_subject_018PJS_wf/anat_preproc_wf/brain_extraction_wf/atropos_wf/01_atropos
Node inputs:
args =
convergence_threshold = 0.0
dimension = 3
environ = {‘NSLOTS’: ‘8’}
icm_use_synchronous_update =
initialization = KMeans
intensity_images = [’/u/project/cparkins/data/kshap_cyberball/work/fmriprep_wf/single_subject_018PJS_wf/anat_preproc_wf/brain_extraction_wf/inu_n4/mapflow/inu_n40/sub-018PJS_T1w_ras_valid_maths_corrected.nii.gz’]
likelihood_model = Gaussian
mask_image = /u/project/cparkins/data/kshap_cyberball/work/fmriprep_wf/single_subject_018PJS_wf/anat_preproc_wf/brain_extraction_wf/get_brainmask/tpl-OASIS30ANTs_res-01_label-brain_probseg_trans_resampled_maths_maths.nii.gz
maximum_number_of_icm_terations =
mrf_radius = [1, 1, 1]
mrf_smoothing_factor = 0.1
n_iterations = 3
num_threads = 8
number_of_tissue_classes = 3
out_classified_image_name =
output_posteriors_name_template = POSTERIOR%02d.nii.gz
posterior_formulation =
prior_probability_images =
prior_probability_threshold =
prior_weighting =
save_posteriors =
use_mixture_model_proportions =
use_random_seed = True
Traceback (most recent call last):
File “/u/project/CCN/apps/fmriprep/1.4.0_py3.7.2/lib/python3.7/site-packages/nipype/pipeline/plugins/multiproc.py”, line 69, in run_node
result[‘result’] = node.run(updatehash=updatehash)
File “/u/project/CCN/apps/fmriprep/1.4.0_py3.7.2/lib/python3.7/site-packages/nipype/pipeline/engine/nodes.py”, line 472, in run
result = self._run_interface(execute=True)
File “/u/project/CCN/apps/fmriprep/1.4.0_py3.7.2/lib/python3.7/site-packages/nipype/pipeline/engine/nodes.py”, line 563, in _run_interface
return self._run_command(execute)
File “/u/project/CCN/apps/fmriprep/1.4.0_py3.7.2/lib/python3.7/site-packages/nipype/pipeline/engine/nodes.py”, line 643, in _run_command
result = self._interface.run(cwd=outdir)
File “/u/project/CCN/apps/fmriprep/1.4.0_py3.7.2/lib/python3.7/site-packages/nipype/interfaces/base/core.py”, line 375, in run
runtime = self._run_interface(runtime)
File “/u/project/CCN/apps/fmriprep/1.4.0_py3.7.2/lib/python3.7/site-packages/nipype/interfaces/ants/segmentation.py”, line 163, in _run_interface
runtime = super(Atropos, self)._run_interface(runtime)
File “/u/project/CCN/apps/fmriprep/1.4.0_py3.7.2/lib/python3.7/site-packages/nipype/interfaces/base/core.py”, line 758, in _run_interface
self.raise_exception(runtime)
File “/u/project/CCN/apps/fmriprep/1.4.0_py3.7.2/lib/python3.7/site-packages/nipype/interfaces/base/core.py”, line 695, in raise_exception
).format(**runtime.dictcopy()))
RuntimeError: Command:
Atropos --image-dimensionality 3 --initialization KMeans[3] --intensity-image /u/project/cparkins/data/kshap_cyberball/work/fmriprep_wf/single_subject_018PJS_wf/anat_preproc_wf/brain_extraction_wf/inu_n4/mapflow/_inu_n40/sub-018PJS_T1w_ras_valid_maths_corrected.nii.gz --likelihood-model Gaussian --mask-image /u/project/cparkins/data/kshap_cyberball/work/fmriprep_wf/single_subject_018PJS_wf/anat_preproc_wf/brain_extraction_wf/get_brainmask/tpl-OASIS30ANTs_res-01_label-brain_probseg_trans_resampled_maths_maths.nii.gz --mrf [0.1,1x1x1] --convergence [3,0] --output [sub-018PJS_T1w_ras_valid_maths_corrected_labeled.nii.gz] --use-random-seed 1
Standard output:
Standard error:
Return code: 1

effigies · December 1, 2019, 1:14pm

I would look at your data to see if there are any obvious issues. You can look at the original T1w image, but also the two specific inputs to this step:

/u/project/cparkins/data/kshap_cyberball/work/fmriprep_wf/single_subject_018PJS_wf/anat_preproc_wf/brain_extraction_wf/inu_n4/mapflow/_inu_n40/sub-018PJS_T1w_ras_valid_maths_corrected.nii.gz
/u/project/cparkins/data/kshap_cyberball/work/fmriprep_wf/single_subject_018PJS_wf/anat_preproc_wf/brain_extraction_wf/get_brainmask/tpl-OASIS30ANTs_res-01_label-brain_probseg_trans_resampled_maths_maths.nii.gz

I would look with something like FreeView or FSLeyes.

ebaek · December 2, 2019, 9:03pm

@effigies Thank you so much for your response. I believe there is something wrong with both the T1w image and the tpl-OASIS30ANTs_res-01_label-brain_probseg_trans_resampled_maths_maths.nii.gz image, with the R/L being flipped in the top left image in the T1w and the A/P being flipped in the second image in the brain mask file. I am guessing I’d need to correct this in the header information and re-run fMRIPrep?

effigies · December 2, 2019, 9:06pm

What viewer are you using? I would use one that will orient according to the header, not the data block.

ebaek · December 2, 2019, 9:17pm

The screenshots above are from fslview, and here are the same images using fsleyes (that I believe still show the problematic orientation?):

effigies · December 2, 2019, 10:00pm

Yup, the A/P flip is definitely there. Not sure how you’re identifying an R/L flip…

Could you open an issue in Niworkflows, where this workflow is implemented?

ebaek · December 2, 2019, 10:31pm

Thanks @effigies! I will open an issue in Niworkflows. I am not sure if my reading/interpretation of the image is correct, but in the middle T1w image, if we are using the subject’s eyes as a guideline then it would seem that “R” indicates the right side of the subject’s brain. Since fsleyes defaults to a radiological display, I thought that this would be incorrect.

effigies · December 2, 2019, 10:46pm

I don’t recall much about how they’re actually displaying it, but it appears to me that it’s trying to be consistent with “Right is Up” between the coronal and axial slices, so coronal is in radiological and axial is in neurological. So it appears inconsistent.

In any event, I would go by the headers, not the display orientation. If FSLeyes is labeling one side R, then that’s (almost certainly) what’s in your image header. If the headers aren’t correct going in, then any preprocessing is compromised. There’s no real way to recover without some additional information, such as knowledge of asymmetries that can be visually judged.