Comparing DRR with x-ray image

I want to compare a DRR of a CT scan with an Xray. Is there any way that I can find an optimal transfer function using VTK ?


CT and X-ray images both use X-rays for acquiring images but they are very different imaging modalities. There is no raycast transfer function that you can apply to a CT to get an image that exactly matches an X-ray:

  • CT-s resolution is very low compared to X-ray, so thin structures will be blurred in the DRR
  • lots of non-linear, spatially varying filtering is applied to X-ray images to optimally utilize dynamic range of the detector or film and reduce artifacts, while clinical CTs are fairly accurately calibrated for radiologic density
  • X-ray imaging uses cone beam, while CT uses pencil beam, therefore X-ray images are much more affected by scattering - which changes the overall background intensity in X-rays images
  • used beam energies are different
  • if the X-ray system uses an image intensifier (typical on older interventional systems, mobile C-arms) then you will have spatial distortions (pincushion and S distortion) in the X-ray image
  • noise characteristics are completely different

If by CT you also mean CBCT then some of the above issues are not relevant (especially if the same device acquired both the CBCT and the projection image) but there are other problems instead (e.g., limited size of the reconstructed CBCT compared to the projection image).

For all these reasons, the DRRs that you create from CTs will only be qualitatively similar to real-world X-ray images. Therefore, if the goal is to generate DRRs that look similar to certain kind of X-ray images then visual assessment by experts should be enough. If the goal is to generate DRRs for 2D/3D registration then you should use similarity metrics that are robust to the above listed differences, for example only rely on local intensity variations (edges, gradients).

VTK forum is mainly about computer graphics and visualization, so if you want to further discuss this topic with more medical imaging research and clinical experts then the Slicer forum or ITK forum are better choices.