[Rtk-users] Problem reconstructing from synthetic data

[Albert MURIENNE]

Hi Simon,

My bad, you are right again when changing the rotation direction everything works fine.
Your indication made me double check my dataset generation code, and there was indeed a little convention mismatch in my new generation code compared to the initial one.

Thanks again and have a nice day,

Albert

De : Simon Rit [mailto:simon.rit at creatis.insa-lyon.fr]
Envoyé : dimanche 20 août 2023 08:32
À : Albert MURIENNE <Albert.MURIENNE at b-com.com>
Cc : rtk-users at openrtk.org
Objet : Re: [Rtk-users] Problem reconstructing from synthetic data

Hi Albert,
It seems that you have changed the rotation direction. Setting -a option of rtksimulatedgeometry to 360 (the default value) seems to fix the problem.
Simon

On Fri, Aug 18, 2023 at 2:10 PM Albert MURIENNE <Albert.MURIENNE at b-com.com<mailto:Albert.MURIENNE at b-com.com>> wrote:
Hello again Simon,

Going further away with my gVirtualXRay/RTK tests, I tried to switch to a new projections set, using the previous indications you gave me.
Unfortunately the results are not very good, and I can’t understand why.
This time the data corresponds to a human phantom, with a 600mm SID and 1200mm SDD, 2.0mm pixel size with a 512x512 image resolution.
I used the following scripts without success:

rtksimulatedgeometry -n 360 -a -360 -o projections_p4_512x512_2.0_600_1200/geometry.xml  --sdd 1200 --sid 600
rtkfdk -p projections_p4_512x512_2.0_600_1200 -r ' attenuation-([0-9]*).tif' -o fdk.mha --neworigin -511.0,-511.0,0 --newspacing 2.0 --nsort --submatch 1 --hardware cuda –g projections_p4_512x512_2.0_600_1200/geometry.xml

This new dataset’s images, as well as the xml geometry I generated, are available for download at this link:

https://filesender.renater.fr/?s=download&token=acf071a0-1299-4725-81d3-4a41b5a8a158

Thanks for your help,

Regards,

Albert

De : Simon Rit [mailto:simon.rit at creatis.insa-lyon.fr<mailto:simon.rit at creatis.insa-lyon.fr>]
Envoyé : lundi 26 juin 2023 11:49
À : Albert MURIENNE <Albert.MURIENNE at b-com.com<mailto:Albert.MURIENNE at b-com.com>>
Cc : rtk-users at openrtk.org<mailto:rtk-users at openrtk.org>
Objet : Re: [Rtk-users] Problem reconstructing from synthetic data

Yes, the origin is the center of the first pixel in ITK see e.g. figure 4.1 of ITK's software guide<https://itk.org/ITKSoftwareGuide/html/Book1/ITKSoftwareGuide-Book1ch4.html#x45-54001r1>.
Simon

On Mon, Jun 26, 2023 at 11:03 AM Albert MURIENNE <Albert.MURIENNE at b-com.com<mailto:Albert.MURIENNE at b-com.com>> wrote:
Hello Simon,

Glad to hear you’re still actively maintaining RTK ☺
I made some tests following your advices, and everything is now working fine!
I had already started playing with the projections origin, but I was missing the point about the rotation convention.
I guess the xxx.75 origin coordinates are related to the origin being at the “center” of the pixel?

Regards,

Albert

De : Simon Rit [mailto:simon.rit at creatis.insa-lyon.fr<mailto:simon.rit at creatis.insa-lyon.fr>]
Envoyé : jeudi 22 juin 2023 11:47
À : Albert MURIENNE <Albert.MURIENNE at b-com.com<mailto:Albert.MURIENNE at b-com.com>>
Cc : rtk-users at openrtk.org<mailto:rtk-users at openrtk.org>
Objet : Re: [Rtk-users] Problem reconstructing from synthetic data

Hi Albert,
Nice to hear that you are still working on CT. There are two things that must be accounted for
- there is a difference in convention for the rotation angle between gVirtualXRay and RTK,
- the projections are not centered if the origin is 0.
The following command lines work for me:
rtksimulatedgeometry -n 360 -a -360 -o g  --sdd 1000 --sid 500
rtkfdk -p . -r 'projection([0-9]*).tif' -o fdk.mha --neworigin -159.75,-119.75,0 --newspacing 0.5 --nsort --submatch 1 --hardware cuda -g g
Let me know if something is unclear,
Simon

On Wed, Jun 21, 2023 at 10:04 AM Albert MURIENNE <Albert.MURIENNE at b-com.com<mailto:Albert.MURIENNE at b-com.com>> wrote:
Hello,

I'm struggling to reconstruct properly a synthetic projection images sequence using RTK.
The sequence was generated using gVirtualXRay (https://gvirtualxray.fpvidal.net/) and a dragon 3D model, and represents a full 360 degrees CBCT tomographic sequence, with a 500mm SID and 1000mm SDD, 0.5mm pixel size with a 640x480 image resolution, and without any offsets (perfect circular geometry). I tried reconstructions with the simplest FDK and SART implementations, inspired by RTK provided code examples, using rtk::ProjectionsReader to read the images and rtk::ThreeDCircularProjectionGeometry to read the geometry, but the results are not good, and I don't really know if the problem comes from the data or from my reconstruction implementation.

The images data, as well as the xml geometry I generated, are available for download at this link:

https://filesender.renater.fr/?s=download&token=bc8f87b4-989c-46d7-8c4b-dac39a4df0e1

Could you help me pinpoint the origin of my problem? Does it come from the data itself?

Thanks,

Albert
Murienne

_______________________________________________
Rtk-users mailing list
rtk-users at openrtk.org<mailto:rtk-users at openrtk.org>
https://www.creatis.insa-lyon.fr/mailman/listinfo/rtk-users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.creatis.insa-lyon.fr/pipermail/rtk-users/attachments/20230822/2abad279/attachment-0001.htm>