VTK merge request: https://gitlab.kitware.com/vtk/vtk/merge_requests/4601

The vtk change has been merged in to vtk master.

@lassoan, could you please try vtk master to see if this works for you?

Thank you, I'll test this today.

Sorry for the delay, I had some problems rebuilding Slicer with latest VTK master, but it works now.

I've tested the volume renderer but unfortunately as soon as I switch to Volume renderer module, all VTK render windows turn to solid gray and stay like that. I could not even test any volume rendering feature. I haven't changed anything else anywhere, just upgraded VTK.

The problem is exactly as described in this bug report: https://issues.slicer.org/view.php?id=4598. It is not clear if there is a solution or workaround that I could apply.

It is not clear if there is a solution or workaround that I could apply

A solution has been identified, I am working on updating CTK to address this.

The easiest is to use QVTKOpenGLNativeWidget every where the QVTKOpenGLWidget is used.

If you do this replacement it should now work.

I've tested this on the latest nightly release. Unfortunately, setting clipping intensity has no effect.

How to test:

• Load CTChest sample data
• In Volume Rendering module, enable GPU volume rendering, clip into the volume
• Run the following commands in the Python console:

<pre>
vpn=slicer.mrmlScene.GetFirstNodeByClass('vtkMRMLVolumePropertyNode')
vp=vpn.GetVolumeProperty()

Set clipping intensity

vp.SetUseClippedVoxelIntensity(1)
vp.SetClippedVoxelIntensity(-5000)

Test that clipping intensity is properly set

print vp.GetUseClippedVoxelIntensity()
print vp.GetClippedVoxelIntensity()

Test that changing volume property has an effect

vp.SetInterpolationTypeToNearest()

Restore original interpolation type

vp.SetInterpolationTypeToLinear()
</pre>

Looking at the implementation, it seems you need to set the SetShade(1) and also call SetClipSurfaceShade(1)
See https://gitlab.kitware.com/vtk/vtk/merge_requests/4601/diffs

May be you could try:

vpn=slicer.mrmlScene.GetFirstNodeByName('VolumeProperty')
vp=vpn.GetVolumeProperty()

vp.ShadeOn();
vp.ClipSurfaceShadeOn();

# Set clipping intensity
vp.SetUseClippedVoxelIntensity(1)
vp.SetClippedVoxelIntensity(-5000)

# Test that clipping intensity is properly set
print vp.GetUseClippedVoxelIntensity()
print vp.GetClippedVoxelIntensity()

# Test that changing volume property has an effect
vp.SetInterpolationTypeToNearest()
# Restore original interpolation type
vp.SetInterpolationTypeToLinear()

I don't see ClipSurfaceShade property anywhere in VTK master. The merge request that you've referred to superseded by https://gitlab.kitware.com/vtk/vtk/merge_requests/4627.

you've referred to superseded by https://gitlab.kitware.com/vtk/vtk/merge_requests/4627.

Good point. This is definitively an issue then.

Are you able to run the associated VTK test on your workstation then ?

I've figured out why it does not work in Slicer:

1. Clipped voxel intensity is ignored if gradient opacity function is set. Slicer always sets gradient opacity function, but usually all of its values are 1.0, so it has no effect.

Potential solution: If gradient opacity only contains 1.0 values (so it is fully opaque), or disabled (using vp.SetDisableGradientOpacity(1)) then clipped voxel intensity should be used. This could be implemented in UseClippedVoxelIntensity(…) function in vtkVolumeShaderComposer.h.

2. If depth peeling is enabled and the renderer contains a VTK box widget then there are random striping artifacts (regions alternating between smooth and non-smooth surface) on the rendered volume surface.

In Slicer, cropping region is usually visualized using VTK box widget, so most would run into this issue when clipping a volume.

Box widget has serious rendering issues. This problem might be related to incorrect rendering box widget when depth peeling is enabled (see 0004609).

If gradient opacity is enabled then the feature does not work at all:

clipping-gradient-opacity-enabled.png (466,555 bytes)

Clipping surface is correct if vtkVolumeShaderComposer.h is changed to take into account GetDisableGradientOpacity() and vp.SetDisableGradientOpacity(1) is called (and depth peeling is disabled).

Modified HasGradientOpacity function:

<pre>
bool HasGradientOpacity(vtkOpenGLGPUVolumeRayCastMapper::VolumeInputMap& inputs)
{
for (auto& item : inputs)
{
vtkVolumeProperty* volProp = item.second.Volume->GetProperty();
const bool gradOp = volProp->HasGradientOpacity() && !volProp->GetDisableGradientOpacity();
if (gradOp)
return true;
}
return false;
}
</pre>

clipping-gradient-opacity-disabled-depth-peeling-off.png (386,157 bytes)

If we enable depth peeling then striping artifacts appear:

clipping-gradient-opacity-disabled-depth-peeling-on.png (418,370 bytes)

Box widget can be anywhere in the render view, it causes artifacts appear on the volume surface:

clipping-depth-peeling-enabled-boxwidget.png (308,646 bytes)

Other software will run into the same issue, so at least the limitation should be clearly documented (https://www.vtk.org/doc/nightly/html/classvtkVolumeProperty.html#a614b3114b709495810c438078fe9c4a6). Maybe even logging a warning would be appropriate when clipped voxel intensity is not used despite UseClippedVoxelIntensity is enabled.

@lassoan,

Sorry for the delay in responding - I was out last week.

When implementing the clipped voxel intensity mode, I chose to leave gradient magnitude calculations as is considering that the magnitude should be "true" magnitude. In hindsight, it seems it would be better to use the clipped voxel intensity for gradient opacity magnitude calculations as well. I'll make the change.

As for depth peeling, I think it is a separate issue. But I'll take a look at it as well.

Great, thank you!

@lassoan, Mind trying out this change:

https://gitlab.kitware.com/sankhesh/vtk/commits/gradient_magnitude_clip

It ensures that the clipped voxel intensity is used for gradient magnitude computations as well.

I tested this and commented at the pull request. It works great except when depth peeling is enabled and there are semi-transparent actors in the scene (anywhere, does not have to be inside or near the volume) - in that case artifacts appear at the clipped surface.

Here is the following up MR from @allison.vacanti
https://gitlab.kitware.com/vtk/vtk/merge_requests/4780

From @lassoan (copied from the MR referenced below): I've spent a few hours testing this and although there were still rendering settings that produced unexpected artifacts (when sampling distances are larger, when cut surface is semi-transparent, or when there are abrupt changes in transfer functions), with some experimentation most of the time I could achieve nice smooth cut surface. I think this is good to go! Thank you very much Allison.

Fixed in r27498
See http://viewvc.slicer.org/viewvc.cgi/Slicer4?view=revision&amp;revision=27498