Closed theysp closed 2 years ago
I don't think the problem is in the dispaced detector weighting. You can validate this with a simulation. The problem is most likely that your input dataset is imperfect (calibration, scatter, etc.) and these imperfections are enhanced by the displaced detector weighting, a known side effect of this strategy. I'm closing the issue but you can re-open it if you still have the issue with simulations.
Thank you, I will try simulation. And let you know if things goes well. Would ART reconstruction suffer less than FDK from the imperfection of data?
It is always hard to fully anticipate but I think that the problem will also be here with iterative algorithms.
Thanks. In another word, image post processing is always mandatory. Correct?
Or adequate pre-processing of the projections
Dear Simon,
I have tried simulation on a CT data of a phantom of chest, generated DRRs using siddon method with 40% offset geometry settings. And reconstructed it using iterative fdk reconstruction. Still the dark hole in the center appears.
As I reviewed a lot of papers, no such obvious artifacts has been reported.
I can share the simulation data where you want. Would you please leave a email address? So I can send the link to the original DRRs as projection images.
I don't think the problem is in the dispaced detector weighting. You can validate this with a simulation. The problem is most likely that your input dataset is imperfect (calibration, scatter, etc.) and these imperfections are enhanced by the displaced detector weighting, a known side effect of this strategy. I'm closing the issue but you can re-open it if you still have the issue with simulations.
I can not re-open this by myself, because you have more privilege on it than me.
Please provide a set of rtk command line to reproduce the issue or a geometry file and one projection header. That should be enough. I'll have a look next week.
Please provide a set of rtk command line to reproduce the issue or a geometry file and one projection header. That should be enough. I'll have a look next week.
Thank you, here are the items you may need:
the cmdline:
rtkiterativefdk.exe --verbose --geometry=_geom.xml --output=rec.tif --hardware=cuda --positivity --pad=0.3 --hann=0.8 --hanY=0.0 --fp=CudaRayCast --path=XXXX --regexp=.*\.tif --lowercrop=4,4,0 --uppercrop=4,4,0 --airthres=32000.0 --iO=65536 --idark=0 --origin=-200,-130,-200 --dimension=400,260,400 --spacing=1.0,1.0,1.0 --direction=1.0,0.0,0.0,0.0,1.0,0.0,0.0,0.0,1.0
the geometry file: _geom.zip
the projection file header:
- dimensions: 896x896
- spacing: 0.483x0.483
- origin: 0.0 0.0
Dear Simon
I have found the reason caused this artifact.
By default, the cuda based iterative fdk reconstructor turns the pad of displacement off. I turn it on by m_DisplacedDetectorFilter->SetPadOnTruncatedSide(true);
And this shadow hole went away.
Would you please confirm this? maybe set the pad on truncated side on by default. If possible, would you please have a simple explain about this phenomenon?
Thank you very much.
Thanks for your tests. The ramp filter works as expected and there is no reason to revise it as far as I know. You can validate it with test straight FDK, not the iterative version. However, I agree with our suggestion. Can you please make it a pull request? If you prefer, I will do it myself.
I'd love to make it a pull request. 10 minutes
Closed with #501
@theysp how to use this filter to do the recon process, may you share your demo?
You used the exe.
I used the following configuration of CBCT device
I tried the iterative FDK reconstruction, set displacement filter on, I got a reconstructed image with a darker central hole. Truncation Correction from 0.0 to 1.0 has been tried as well as various Air Threshold for the averaged scatter preprocessing and Hann Cut Frequency. The darker central hole varies from dark hole to a not so dark hole with a shining halo around it.
Since the code of displacement has not been changed for a long time, could it be improved if I use other weighting strategy?
Thank you very much.
update on 2022-06-29 I have tried simulation on a CT data of a phantom of chest, generated DRRs using siddon method with 40% offset geometry settings. And reconstructed it using iterative fdk reconstruction. Still the dark hole in the center appears.
As I reviewed a lot of papers, no such obvious artifacts has been reported.