is there a way to specify dimensions in CropForegroundd ?

[scarpma]

I mean if I want to crop a volume only along the superior-inferior direction by using CropForeground function. I don't want to crop also along R-L and A-P directions.

I thought that setting select_fn arg to lambda x: np.all(x>0, axis=3) would do the job, but an error is raised

Original Traceback (most recent call last):
  File "/opt/miniconda3/envs/pytorch3d/lib/python3.8/site-packages/monai/transforms/transform.py", line 89, in apply_transform
    return _apply_transform(transform, data, unpack_items)
  File "/opt/miniconda3/envs/pytorch3d/lib/python3.8/site-packages/monai/transforms/transform.py", line 53, in _apply_transform
    return transform(parameters)
  File "/opt/miniconda3/envs/pytorch3d/lib/python3.8/site-packages/monai/transforms/croppad/dictionary.py", line 884, in __call__
    d[key] = self.cropper.crop_pad(img=d[key], box_start=box_start, box_end=box_end, mode=m)
  File "/opt/miniconda3/envs/pytorch3d/lib/python3.8/site-packages/monai/transforms/croppad/array.py", line 756, in crop_pad
    pad_to_end = np.maximum(box_end - np.asarray(img.shape[1:]), 0)
ValueError: operands could not be broadcast together with shapes (2,) (3,)

[scarpma]

Ok, I found a solution, but I have another problem.

The solution is to use a select_fn of this kind

def crop_only_axially(array):
    idx_bool = np.any(np.any(array>0, axis=-2), axis=-2)
    return idx_bool.expand_as(array)

it assumes that the input is in RAS coordinate system.

The other problem I found is that CropForegroundd does not update the affine matrix. So I cannot save data back to nifti format (it would be wrong).

Actually there is an open PR #3681 that already implemented the needed changes, but it still lacks of one check. It seems stuck from January, tough. Does anyone know if someone is still working on this ?

Thanks !

[scarpma]

I found a solution. See this PR for info.