Interpolation of Resample class

[kodohoi]

When I used Resample class to change the spacial infromation of CT images, my CT images looked distorted on 3D view in 3D slicer. Is this problem caused by linear interpolation which is the default interpolation of Torchio? I confirmed the affine of CT images changed after transformation of Resample.

[fepegar]

Can you explain what you run, what you expected and share a screenshot of what you got?

[fepegar]

Thanks. Can you please share a minimal example and one image to try it on? Also, it might be more useful to show us the 2D views in Slicer, instead if a posterior 3D view. I think ChatGPT lacks some context in this case. It might think you're referring to intensity distortion, as opposed to geometric distortions.

A bit off topic, but you could instead open the PDF of the TorchIO documentation (https://torchio.readthedocs.io/_/downloads/en/latest/pdf/) in Edge, ask the Bing Chat to read it, and ask questions.

[kodohoi]

Thank you.
ZA-torchio_000.nii.gz
ZA-torchio_000.zip
I share the DICOM images before processing and processed nifti image.

[kodohoi]

@fepegar
I apologize for the inconvinience.
I think this geometoric distortion of my data may be caused by Gantry/Detector Tilt.
I guess nifti files do not contain the information about the tilt angle of a CT, and this result in the geometirc distortion...
I am very sorry...

Does the tilt angle affect the affine matrix of image?
If so, is it possible to deal with the effect of the tilt angle with Torchio?

[fepegar]

No worries. Yes, the tilt angle will affect the affine. But it's probably stored in the NIfTI sform matrix, because the qform can only store a rigid transformation. You could resample your image using the sform so the tilting is not visible in your voxel data.

[fepegar]

I recommend you fix your images as explained in the Slicer forum: https://discourse.slicer.org/t/gantry-tilt-correction-in-ct-dicoms/19443