Spacing transform does not update transformed image meta dictionary?

[agaldran]

Hello,

Sorry about the naive question, I am trying to wrap my head around transforms and meta_dicts. Right now I have the following operations:

import monai.transforms as t

data_dict = {'image': 'data/ct/images/.whatever.nii.gz'}
loader = t.LoadImaged(keys=('image'), image_only=False)
loaded_data_dict = loader(data_dict)
print(loaded_data_dict['image'].shape, loaded_data_dict['image_meta_dict']['pixdim'][1:4])

which returns a shape and voxel sizes of [281, 313, 182] and (0.50,0.50,0.75) respectively. Now, if I run:

spacing = t.Spacingd(keys=['image'], pixdim=(0.50, 0.50, 1.5), mode=('bilinear')) 
spaced_data_dict = spacing(loaded_data_dict)
print(spaced_data_dict['image'].shape, spaced_data_dict['image_meta_dict']['pixdim'][1:4])

I am expecting to see approx half the voxels in the Z direction, as the original spacing was 0.75 and I am asking for twice that spacing. For the other directions, nothing should be changed^* . This is what I get:

[281, 313, 92], (0.50,0.50,0.75)

So, although the number of voxels are as expected, the spacing remains the same. I cannot find any other key in the meta data dictionary that tells me what is the current spacing for this volume. Am I doing something wrong here?

^* not sure about this either but numbers seem to match. Is my understanding correct? It also happens that if I use (0.0, 0.0, 1.5) I get the same result as when I specify the original voxel sizes, is this expected behavior?

[KumoLiu]

Hi @agaldran, thanks for your interest here.

• First, pixdim should also be updated after Spacing, I think it's a bug here, so I will create an issue for it.
• You can also use spaced_data_dict['image_meta_dict']["affine"] to get the new affine which contains the pixdim information.
• When you use non-positive values, the transform will use the corresponding components of the original pixdim, which is computed from the affine matrix of the input image, so it will return the original voxel size, which is expected.
  

  MONAI/monai/transforms/spatial/array.py

  Lines 362 to 369 in cb257d2

  	pixdim: output voxel spacing. if providing a single number, will use it for the first dimension.
  	items of the pixdim sequence map to the spatial dimensions of input image, if length
  	of pixdim sequence is longer than image spatial dimensions, will ignore the longer part,
  	if shorter, will pad with the last value. For example, for 3D image if pixdim is [1.0, 2.0] it
  	will be padded to [1.0, 2.0, 2.0]
  	if the components of the `pixdim` are non-positive values, the transform will use the
  	corresponding components of the original pixdim, which is computed from the `affine`
  	matrix of input image.

BTW, you can use spaced_data_dict["image"].meta instead of spaced_data_dict['image_meta_dict'] to get meta information directly after monai v0.9.1.
https://github.com/Project-MONAI/MONAI/wiki/MetaTensor-guide
Hope it helps, thanks!

[agaldran]

That is very interesting thanks a lot!

Expect me to be back in this Q&A in the next weeks, I am finally having some time to dig in MONAI code and I my hat is full of questions :)

Cheers,

Adrian