`n4_correct` from `early_b0ref_wf` fails on some subjects

[josephmje]

Description

N4BiasFieldCorrections fails on a subset of participants in one dataset that I've been testing. The subjects were all scanned on a Siemens Prisma.

What I Did

N4BiasFieldCorrection \
  --bspline-fitting [ 200 ] -d 3 \
  --input-image /output/sub-CMPP998/early_b0ref_wf/dwi_reference_wf/enhance_and_skullstrip_dwi_wf/n4_correct/sub-CMPP998_ses-01_dwi_b0_mcf_ref_scaled.nii.gz \
  --mask-image /output/sub-CMPP998/early_b0ref_wf/dwi_reference_wf/pre_mask/sub-CMPP998_ses-01_dwi_b0_mcf_mask.nii.gz \
  --output sub-CMPP998_ses-01_dwi_b0_mcf_ref_corrected.nii.gz -r

When running with the --verbose flag, I get the following error:

Description: itk::ERROR: N4BiasFieldCorrectionImageFilter(0x2c89620): Inputs do not occupy the same physical space!
InputImage Spacing: [8.0000000e+00, 8.0000000e+00, 8.0000000e+00], InputImage_1 Spacing: [8.0000000e+0              0, 7.9999995e+00, 7.9999909e+00]
        Tolerance: 8.0000000e-06

[oesteban]

@josephmje please have a look to nipreps/niworkflows#600 - much of what you are saying here will definitely surface on that undertaking.

Can you share some data? (or be ready to test the new developments as they arrive??)

LMKWYT

EDIT: I actually meant nipreps/niworkflows#601

[josephmje]

@oesteban sorry for the late reply on this. i'd be happy to share some data. do you have a preferred method for receiving them?

[oesteban]

#169 should've addressed this - could you test @josephmje ?

[celprov]

@oesteban I'm running into the same error with our hcph-pilot data, a deep sampling dataset with 16 sessions on a single individual.

At first, I wondered whether this error was a consequence of running only the anatomical workflow and then the dwi workflow. But given that @josephmje seemed to face it with subjects also scanned on a Siemens Prisma, maybe it's a problem of dMRIPrep being unable to handle data from this scanner.