[Rtk-users] Support for translate-rotate geometries with small FOV detectors

[Simon Rit]

Hi Sean,
Interesting! Briefly,
- for rtkfdk, there is nothing ready out of the box. The displaced detector
only works for truncated projections on one side only. One should be able
to calculate weights for your case, weighting both sides of the projections
except except for the furthest one from the center (which should not be
truncated on one side).
- for iterative algorithms, we have implemented the use of the displaced
weights following e.g. http://dx.doi.org/10.1088/0031-9155/58/2/205 but you
can disable them with the --nodisplaced option. There is no theoretical
ground for having them except avoiding creating discontinuities so that the
algorithm converges to a smooth result in the central overlap region.
I hope it helps,
Simon

On Tue, Jun 24, 2025 at 7:26 AM Sean Phillip Hood <sph864 at uowmail.edu.au>
wrote:

> Hi,
>
> I am using RTK to reconstruct spectral CBCT images from projections
> obtained with a photon counting detector (PCD) mounted on an Elekta linac.
>
> The main challenge is the size of the PCD (at isocentre the FOV is
> approximately 80 mm x 10 mm). To address this we repeat CBCT acquisitions
> with the detector offset by multiples of its width for each gantry rotation.
>
> Unfortunately, the sampling of gantry angles varies between acquisitions,
> and so the projections cannot be simply stitched together. Furthermore,
> obtaining projections by scanning across the FOV at a series of gantry
> angles is far too slow.
>
> Therefore, I wanted to see if there was a solution using RTK. The first
> thing I have tried is to feed the command line applications all of the
> sorted PCD projections from separate acquisitions as well as a concatenated
> geometry file. The concatenation involves extracting the geometry
> information from each Elekta XML file and appending it to a single geometry
> file. The translation is added to the projection offset in the x-direction.
>
> For RTK v2.5.0, this results in the following exception thrown by
> rtkDisplacedDetectorImageFilter when using rtkadmmtotalvariation:
>
> Description: ITK ERROR: Cannot account for detector displacement larger
> than 50% of panel size. Corner inf=59.661 and corner sup=682.68
>
> A similar error is returned by rtkfdk. It seems that projection offsets
> larger than 50% of the detector cannot be used. Would it be possible to add
> support for this, or for translate-rotate geometries in general?
>
> I only ask because I have also tried this method with an older version of
> RTK (I do not know the exact version since I was provided with pre-compiled
> executables). These errors did not occur and I was able to at least produce
> an output from rtkadmmtotalvariation, albeit one that had clear geometry
> issues (I suspect there are differences in definitions that I'm not
> accounting for between RTKThreeDCircularGeometry versions 2 and 3, the
> former being used by the older version of RTK).
>
> Please let me know if I can provide any more info, and I would be grateful
> for any assistance. Thanks!
>
> Kind regards,
> *Sean Hood*
> School of Physics, Faculty of Engineering and Information Sciences
> University of Wollongong, Australia
> _______________________________________________
> Rtk-users mailing list
> 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/20250624/923c7249/attachment.htm>