ITK ERROR: ExtractImageFilter

[jpmcgeown]

I'm encountering an ITK bug with the current version of ANTsPyMM. The traceback is provided below. Package versions are as follows:
{'tensorflow': '2.20.0', 'antspyx': '0.6.2', 'antspynet': '0.3.2', 'antspyt1w': '1.1.3', 'antspymm': '1.6.4'}. I've tried building from source and installing via pip and I reproduce the error in both instances.

I've confirmed the issue isn't with the test data as it runs without issue in an old version of ANTsPYMM with these libraries: {'tensorflow': '2.17.0', 'antspyx': '0.4.2', 'antspynet': '0.2.8', 'antspyt1w': '0.9.4', 'antspymm': '1.4.0'}.

I'm setting up to run ANTsPyMM on 600+ datasets in the coming weeks, across two HPC systems. I would, ideally, prefer to install and run the most recent version of ANTsPyMM on both systems rather than rebuilding the old version.

Any help would be appreciated, let me know if you need anymore information.

Josh

RUN sub-109 --- /nesi/nobackup/uoa04671/projects/antspymm_test/antspymm4
{'tensorflow': '2.20.0', 'antspyx': '0.6.2', 'antspynet': '0.3.2', 'antspyt1w': '1.1.3', 'antspymm': '1.6.4'}
--:
False False False False
Read
bxt
qc results
Begin registration
Begin Atropos tissue segmentation
do deep_tissue_segmentation
End Atropos tissue segmentation
Begin DKT
hippocampus
medial temporal lobe
deep CIT168
SN-specific segmentation
Traceback (most recent call last):
File "/nesi/project/uoa04671/jmcg465/software-tools/projects/simlr/scripts/02_job_script.py", line 90, in
mmrun = antspymm.mm_csv(studycsv2,
^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/nesi/project/uoa04671/jmcg465/conda_envs/antspymm_build/lib/python3.12/site-packages/antspymm/mm.py", line 9026, in mm_csv
hier = antspyt1w.hierarchical( t1, hierfn, labels_to_register=None )
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/nesi/project/uoa04671/jmcg465/conda_envs/antspymm_build/lib/python3.12/site-packages/antspyt1w/get_data.py", line 2881, in hierarchical
snreg = region_reg( img, myparc['tissue_segmentation'], ibinseg,
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/nesi/project/uoa04671/jmcg465/conda_envs/antspymm_build/lib/python3.12/site-packages/antspyt1w/get_data.py", line 1791, in region_reg
synL = localsyn(
^^^^^^^^^
File "/nesi/project/uoa04671/jmcg465/conda_envs/antspymm_build/lib/python3.12/site-packages/antspyt1w/get_data.py", line 1592, in localsyn
tcrop = ants.crop_image(themi, hemicropmask)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/nesi/project/uoa04671/jmcg465/conda_envs/antspymm_build/lib/python3.12/site-packages/ants/ops/crop_image.py", line 60, in crop_image
itkimage = libfn(image.pointer, label_image.pointer, label, 0, [], [])
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: /project/itksource/Modules/Core/Common/include/itkExtractImageFilter.hxx:91:
ITK ERROR: ExtractImageFilter(0x36fae440): The number of zero sized dimensions in the input image Extraction Region
is not consistent with the dimensionality of the output image.
Expected the extraction region size ([0, 0, 0]) to contain 0 zero sized dimensions to collapse.

[stnava]

this bug arose in antspy and is addressed by ANTsX/ANTsPy@e9e36e4

the logical operation was returning an unsigned char boolean image which is incompatible with how that image is treated quantitatively later in pipelines.

this is more of a perhaps unspoken standard ( float ) for antspy-related processing pipelines. specifically

img != 0

returns a boolean uchar image instead of a float image. would only happen if binarize=True

just fyi: @cookpa @ntustison -- perhaps trigger a new antspy release (0.6.3) ?

certainly possible this type of issue may be around elsewhere but I've not seen it

[jpmcgeown]

Just confirming this patch worked as expected. Thanks for this. I'll keep an eye out for antspy 0.6.3.

[cookpa]

Thanks @stnava, I'll fix ANTsX/ANTsPy#891 and then make a release

[cookpa]

https://pypi.org/project/antspyx/0.6.3/