nnDetection: [Bug] ITK ERROR: ITK only supports orthonormal direction cosines. No orthonormal definition found! (SimpleITK >= 2.1.0)

Hello everyone,

Thanks for this great tool 😃

💀 Bug

When using nndet_prep [task] --full-check command, the following error occured:

File "<nnDetection>/nndet/utils/check.py", line 213, in _full_check
	img_itk_seq = [load_sitk(cp) for cp in case_paths]
File "<nnDetection>/nndet/utils/check.py", line 213, in <listcomp>
	img_itk_seq = [load_sitk(cp) for cp in case_paths]
File "<nnDetection>/nndet/io/itk.py", line 107, in load_sitk
	return sitk.ReadImage(str(path), **kwargs)
File "<CondaEnv>/lib/python3.9/site-packages/SimpleITK/extra.py", line 346, in ReadImage
	return reader.Execute()
File "<CondaEnv>/lib/python3.9/site-packages/SimpleITK/SimpleITK.py", line 8015, in Execute
	return _SimpleITK.ImageFileReader_Execute(self)
RuntimeError: Exception thrown in SimpleITK ImageFileReader_Execute: /tmp/SimpleITK-build/ITK/Modules/IO/NIFTI/src/itkNiftiImageIO.cxx:1980:
ITK ERROR: ITK only supports orthonormal direction cosines.  No orthonormal definition found!

This was noticed with the nnUNet framework when mixing different versions of SimpleITK an issue was raised (https://github.com/SimpleITK/SimpleITK/issues/1433).

When using SimpleITK 2.0.2, this error does not occur. This seems to be due to recent changes with ITK when handling Nifti headers (see https://github.com/InsightSoftwareConsortium/ITK/issues/2674 for further details and ongoing conversation).

I would think that freezing SimpleITK to SimpleITK < 2.1.0 would temporarily solve the issue.

Best

Environment

Environment Information:

----- PyTorch Information -----
PyTorch Version: 1.9.0
PyTorch Debug: False
PyTorch CUDA: 10.2
PyTorch Backend cudnn: 7605
PyTorch CUDA Arch List: ['sm_37', 'sm_50', 'sm_60', 'sm_61', 'sm_70', 'sm_75', 'compute_37']
PyTorch Current Device Capability: (7, 0)
PyTorch CUDA available: True

 


----- System Information -----
System NVCC: nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2019 NVIDIA Corporation
Built on Sun_Jul_28_19:07:16_PDT_2019
Cuda compilation tools, release 10.1, V10.1.243

 

System Arch List: None
System OMP_NUM_THREADS: 1
System CUDA_HOME is None: True
System CPU Count: 6
Python Version: 3.9.6 (default, Jul 30 2021, 16:35:19)
[GCC 7.5.0]

 


----- nnDetection Information -----
det_num_threads 12
det_data is set True
det_models is set True```