Segmentation.to_label_image() issue when specifying ROI name

[JoshuaSiraj]

When I try something like:

mask_scan = read_dicom_auto(mask_path)
mask = mask_scan.to_segmentation(example_image, ["GTVp"]).to_label_image()

I get the following error:

{
	"name": "RuntimeError",
	"message": "Exception thrown in SimpleITK Image_CopyInformation: /tmp/SimpleITK/Code/Common/src/sitkImage.cxx:348:
sitk::ERROR: Source Image for information does not match this image's dimension.",
	"stack": "---------------------------------------------------------------------------
RuntimeError                              Traceback (most recent call last)
/tmp/ipykernel_34497/2353237833.py in ?()
      1 mask_scan = read_dicom_auto(mask_path)
----> 2 mask = mask_scan.to_segmentation(example_image, [\"GTVp\",]).to_label_image()
      3 
      4 visualize_3d_slices_adaptive(mask, axis=0, step=5, cmap=None)

~/Documents/BHKLAB/med-imagetools/src/imgtools/modules/segmentation.py in ?(self)
    324             The label image with each voxel assigned the label of the ROI with the highest value.
    325         \"\"\"
    326         arr, *_ = image_to_array(self)
    327         label_arr = np.where(arr.sum(-1) != 0, arr.argmax(-1) + 1, 0)
--> 328         label_img = array_to_image(label_arr, reference_image=self)
    329         return label_img

~/Documents/BHKLAB/med-imagetools/src/imgtools/utils/imageutils.py in ?(array, origin, direction, spacing, reference_image)
     36         The resulting SimpleITK image.
     37     \"\"\"
     38     image = sitk.GetImageFromArray(array)
     39     if reference_image is not None:
---> 40         image.CopyInformation(reference_image)
     41     else:
     42         image.SetOrigin(origin)
     43         image.SetDirection(direction)

~/Documents/BHKLAB/med-imagetools/.pixi/envs/dev/lib/python3.12/site-packages/SimpleITK/SimpleITK.py in ?(self, srcImage)
   3462         exception will be generated.
   3463 
   3464 
   3465         \"\"\"
-> 3466         return _SimpleITK.Image_CopyInformation(self, srcImage)

RuntimeError: Exception thrown in SimpleITK Image_CopyInformation: /tmp/SimpleITK/Code/Common/src/sitkImage.cxx:348:
sitk::ERROR: Source Image for information does not match this image's dimension."
}

When I specify multiple roi_names like:

mask = mask_scan.to_segmentation(example_image, ["GTVp", "Mandible_Bone"]).to_label_image()

I get an empty label image.

I do get a proper mask when I don't specify rois:

mask = mask_scan.to_segmentation(example_image).to_label_image()

[JoshuaSiraj]

Interestingly, this produces an empty mask:

mask = mask_scan.to_segmentation(example_image, roi_names=['Mandible_Bone'])

And this does not:

mask = mask_scan.to_segmentation(example_image, roi_names={"Man":"Mandible_Bone.*"})

[jjjermiah]

if i was to refactor to_label_image what exactly should it do ?

would be helpful to create some test cases to make sure

[JoshuaSiraj]

if i was to refactor to_label_image what exactly should it do ?

would be helpful to create some test cases to make sure

Currently mask = mask_scan.to_segmentation(example_image) returns a mask with a channels dimension.
to_label_image returns that without a channels dimensions and all the rois in one 3d image.