ENH: MIP image filter for DRR image calculation

[MichaelColonel]

MIPForwardProjectionImageFilter is derived from JosephForwardProjectionImageFilter and computes maximum intensity step along the ray casting on the projection

Example with parallel geometry beam.

DRR image with JosephForwardProjectionImageFilter:

DRR image with MIPForwardProjectionImageFilter:

[SimonRit]

Thank you for your contribution. Have you tried to avoid copy pasting the original code of Joseph's projector? It should not be necessary if one uses a functor.

[MichaelColonel]

I've tried, but couldn't make it. Mainly because the SumAlongRay functor is a tally, in case of MIP we need a step with the maximum intensity not the sum of the steps along a ray. IMHO.

[SimonRit]

I see. I guess including the sum as a reference parameter of SumAlongRay and doing the sum in the functor would solve this? This code also needs to be changed using lambdas in my opinion but that's a more ambitious PR.

[SimonRit]

Or doing a new AccumulateAlongRay lambda?

[MichaelColonel]

I guess including the sum as a reference parameter of SumAlongRay and doing the sum in the functor would solve this?

I will try to use your suggestion, and use sum as a reference parameter in functor.

[MichaelColonel]

Functor is changed, so copy-pasted part is removed.

This code also needs to be changed using lambdas in my opinion but that's a more ambitious PR.

Do you suggest of using lambdas instead of functors?

[SimonRit]

Functor is changed, so copy-pasted part is removed.

Thanks! Looks better.

Do you suggest of using lambdas instead of functors?

Yes, something like what's been done here for ITK unary image filters. It would make the code look nicer but I'm not sure it would be as efficient. To be tested...

[SimonRit]

There are some style errors: https://github.com/RTKConsortium/RTK/actions/runs/7742228424/job/21110871876?pr=579

I would also suggest to change name from MIPForwardProjectionImageFilter to MaximumIntensityProjectionImageFilter to follow ITK's guidelines to avoid abbreviations.

[MichaelColonel]

There are some style errors: https://github.com/RTKConsortium/RTK/actions/runs/7742228424/job/21110871876?pr=579

I would also suggest to change name from MIPForwardProjectionImageFilter to MaximumIntensityProjectionImageFilter to follow ITK's guidelines to avoid abbreviations.

I will fix these errors.

[MichaelColonel]

Filter was renamed, and MIP filter was added to rtkforwardprojections application.

[SimonRit]

I think the code is ready to be merged. It would be good to add a test and the wrapping. If you are not willing to do them, please rebase on master and squash all commits in one. I'll then merge and do them in another PR.

[MichaelColonel]

I can try. I'll let you know in case of problems.

[MichaelColonel]

Is there a instruction how to test python wrapping?

[LucasGandel]

Is there a instruction how to test python wrapping?

Ideally you can add a test similar to rtkFirstReconstruction.py and enable it here. Then you need to build RTK with testing and python wrapping enabled to be able to run the test with ctest

[SimonRit]

Thanks @LucasGandel! I had forgotten about this. @MichaelColonel If you just want to test that newly created wheels will contain the wrapping, wheels are automatically generated by the CI, you can download them as artifacts and test them locally. Currently, this PR does not compile the wheel https://github.com/RTKConsortium/RTK/pull/579/checks#step:5:2085

[MichaelColonel]

I have a question about writing a wrapper for MIP filter. I've taken as a basis the JosephForwardProjectionAttenuated wrapper, since both class are derived from JosephForwardProjection.

My wrapper is

set(WRAPPER_AUTO_INCLUDE_HEADERS OFF)
itk_wrap_named_class("rtk::Functor::MaximumIntensityAlongRay" "rtkFunctorMaximumIntensityAlongRay")
  foreach(t ${WRAP_ITK_REAL})
    itk_wrap_template("${ITKM_${t}}${ITKM_${t}}" "${ITKT_${t}}, ${ITKT_${t}}")
  endforeach()
itk_end_wrap_class()
itk_wrap_named_class("rtk::Functor::MaximumIntensityProjectedValueAccumulation" "rtkFunctorMaximumIntensityProjectedValueAccumulation")
  foreach(t ${WRAP_ITK_REAL})
    itk_wrap_template("${ITKM_${t}}${ITKM_${t}}" "${ITKT_${t}}, ${ITKT_${t}}")
  endforeach()
itk_end_wrap_class()
set(WRAPPER_AUTO_INCLUDE_HEADERS ON)

itk_wrap_class("rtk::JosephForwardProjectionImageFilter" POINTER)
  foreach(t ${WRAP_ITK_REAL})
    itk_wrap_template("I${ITKM_${t}}3I${ITKM_${t}}3SWM${ITKM_${t}}D${ITKM_${t}}IPC"
      "itk::Image<${ITKT_${t}}, 3>, itk::Image< ${ITKT_${t}}, 3>, rtk::Functor::InterpolationWeightMultiplication<${ITKT_${t}}, ${ITKT_${t}}, ${ITKT_${t}}>, rtk::Functor::MaximumIntensityProjectedValueAccumulation<${ITKT_${t}}, ${ITKT_${t}}>, rtk::Functor::MaximumIntensityAlongRay<${ITKT_${t}}, ${ITKT_${t}}>")
  endforeach()
itk_end_wrap_class()

itk_wrap_class("rtk::MaximumIntensityProjectionImageFilter" POINTER)
  foreach(t ${WRAP_ITK_REAL})
    itk_wrap_template("I${ITKM_${t}}3I${ITKM_${t}}3SWM${ITKM_${t}}D${ITKM_${t}}IPC"
      "itk::Image<${ITKT_${t}}, 3>, itk::Image< ${ITKT_${t}}, 3>, rtk::Functor::InterpolationWeightMultiplication<${ITKT_${t}}, ${ITKT_${t}}, ${ITKT_${t}}>, rtk::Functor::MaximumIntensityProjectedValueAccumulation<${ITKT_${t}}, ${ITKT_${t}}>, rtk::Functor::MaximumIntensityAlongRay<${ITKT_${t}}, ${ITKT_${t}}>")
  endforeach()
itk_end_wrap_class()

itk_wrap_class("rtk::MaximumIntensityProjectionImageFilter" POINTER)
  foreach(t ${WRAP_ITK_REAL})
    itk_wrap_template("I${ITKM_${t}}3I${ITKM_${t}}3" "itk::Image<${ITKT_${t}}, 3>, itk::Image<${ITKT_${t}}, 3>")
  endforeach()
itk_end_wrap_class()

During the compilation there are multiple warning messages like below, as a result python packages can't be created.

rtkJosephForwardProjectionImageFilterIF3IF3SWMFDFIPC_Pointer in /work/ITK-cp39-cp39-manylinux_2_28_x64/Wrapping/Typedefs/rtkMaximumIntensityProjectionImageFilter.idx is already defined in /work/ITK-cp39-cp39-manylinux_2_28_x64/Wrapping/Typedefs/rtkJosephForwardAttenuatedProjectionImageFilter.idx.

Do you have a tutorial of some kind, how to write a wrapper for derived classes?

[SimonRit]

The available tutorial is the ITK software guide section 9.5. The warning indicate that you already have an instance with the same name so you need to change the first paramter of the itk_wrap_template macro.

[MichaelColonel]

It looks like compilation with python and tests are OK! Should i squash the commits into one?

[SimonRit]

Yes please squash! I tested the wrapping and it works perfectly. Thanks!

[MichaelColonel]

Squashed in one commit.

[SimonRit]

Thanks for your contribution!