I am trying to copy all the metadata/header information from one image to another.

I have a segmentation (original_image) that when I overlap it with the MRI image it comes from it overlaps as expected (comparing it in 3D Slicer, for instance). I perform some operations on Seg3D (or any other software that saves just the image array) and when I save the image (new_image) the overlap doesn’t occur anymore, it’s moved and rotated. I have the function below using SimpleITK (sitk) to try to copy all the possible metadata/header/information from the original_image but it still doesn’t align. The center seems to be fine, but that’s it. I understand that there are probably redundancies in the script, I was trying all possible options.

def save_itk_keeping_header(new_image, original_image, filename):



  [image_bad_header_itk.SetMetaData(key,image_good_header.GetMetaData(key)) for key in image_good_header.GetMetaDataKeys()]

  sitk.WriteImage(image_bad_header_itk, filename, True)

The output I get is below - it seems that it should be rotated 90 degrees. Any idea what am I missing? In case it helps, the grey image is an MRI, so they are usually rotated in contrast with CT images that are align with the XYZ coordinate system. Also, I know I’m not copying the spacing - it’s one of the modifications I do - but even if I add it, that doesn’t solve my problem.


Hello @Cristobal_Rodero,

Just guessing here, but possibly the coordinate systems were corrupted when you did your work. Try forcing to the standard ITK configuration:

image_good_header = sitk.DICOMOrient(image_bad_header_itk, "LPS")

If this doesn’t work, you will need to share data so that we can reproduce and try to help you (anonymize before sharing or even better, share a small artificial example).

1 Like

Hello @zivy,
Apologies for the late reply. I’ve been doing some more tests and I confirm that this works every time. thank you so much!