[Rtk-users] FDK for planar ct

Cyril Mory cyril.mory at creatis.insa-lyon.fr
Wed Oct 25 03:34:04 EDT 2017


Dear Robert,

Could you send two datasets of projections, one for each case ? We would 
have a look at them and it would help us understand your trajectories. 
The drawings you sent do not seem to be sufficient to remove all 
ambiguities.

Best regards,

Cyril


On 25/10/2017 09:03, "Robert Calließ" wrote:
> Hello,
> the object is moving on a circular path. There are arrows between the 
> different detector positions showing the
> moving direction. The source is static.
> Kind regards,
> Robert
> *Gesendet:* Dienstag, 24. Oktober 2017 um 16:47 Uhr
> *Von:* "Simon Rit" <simon.rit at creatis.insa-lyon.fr>
> *An:* "Robert Calließ" <Robert.Calliess at gmx.de>
> *Cc:* "rtk-users at public.kitware.com" <rtk-users at public.kitware.com>
> *Betreff:* Re: Re: [Rtk-users] FDK for planar ct
> Hi,
> I see one drawing only, not two. And the object does not seem to be 
> moving on your drawing, is it? If not and the source is also static 
> (as it seem), this is equivalent to one large projection.
> Simon
> On Tue, Oct 24, 2017 at 3:58 PM, "Robert Calließ" 
> <Robert.Calliess at gmx.de <mailto:Robert.Calliess at gmx.de>> wrote:
>
>     Hello,
>     I suppose there are still misunderstandings with respect to the
>     trajectory.
>     Attached you can find the two difference trajectories. I also had
>     a closer look to
>     the off centered fdk ( the paper you suggested). But I don't think
>     it is in my case.
>     The iso ray passes object center and detector center at each view.
>     Off centered fdk
>     has a different preweighting scheme.
>     You said that the RTK ramp filter is along the u axis (orthogonal
>     to rotaion axis). For planar_ct_1 trajectory that
>     should fit. As you can see at the picture, the object is moving on
>     a circular path but not rotating around the
>     center point (red cross in the image).
>     Kind regards,
>     Robert C.
>     *Gesendet:* Donnerstag, 12. Oktober 2017 um 07:14 Uhr
>     *Von:* "Simon Rit" <simon.rit at creatis.insa-lyon.fr
>     <mailto:simon.rit at creatis.insa-lyon.fr>>
>     *An:* "Robert Calliess" <robert.calliess at gmx.de
>     <mailto:robert.calliess at gmx.de>>
>     *Cc:* "rtk-users at public.kitware.com
>     <mailto:rtk-users at public.kitware.com>"
>     <rtk-users at public.kitware.com <mailto:rtk-users at public.kitware.com>>
>
>     *Betreff:* Re: [Rtk-users] FDK for planar ct
>     Hello,
>     No. The filter should be orthogonal to the rotation axis.  The RTK
>     ramp filter is along the u axis of the projection.
>     Trajectory 2: if you take photos by rotating the cameras, they are
>     photographies of the same point-of-view. This is what I meant.
>     Cheers,
>     Simon
>     On Wed, Oct 11, 2017 at 8:58 PM, Robert Calliess
>     <robert.calliess at gmx.de <mailto:robert.calliess at gmx.de>> wrote:
>
>         Hello,
>
>         thanks for the link to the paper but I dont have access to it.
>         Aside from how the trajectory is interpreted within in RTK. My
>         actual question was
>
>         if any of those two trajectories would need another
>         reconstruction filter than the FDK Filter. From my point of
>         understanding a specific rotation around
>
>         the object is necessary for fbp/fdk (like c-arm bow, standard
>         circular cone-beam trajectory).  That’s why I asked If the
>         first trajectory needs some other reconstruction
>
>         filter because the object itself doesn’t rotate around itself.
>         It actually gets translated on a circular path. So I was more
>         expecting a “yes” or “no” to the fdk filter
>
>         or a hint to another filter (except iterative reconstructions)
>         I should use for these trajectories.
>
>         To trajectory 2: I think the projections are different. The
>         object rotates and each projection shows a different view.
>
>         Kind regards,
>
>         Robert C.
>
>         *Von:*simon.rit at gmail.com <mailto:simon.rit at gmail.com>
>         [mailto:simon.rit at gmail.com <mailto:simon.rit at gmail.com>] *Im
>         Auftrag von *Simon Rit
>         *Gesendet:* Dienstag, 10. Oktober 2017 20:29
>         *An:* Robert Calliess
>
>
>         *Cc:* Cyril Mory; rtk-users at public.kitware.com
>         <mailto:rtk-users at public.kitware.com>
>         *Betreff:* Re: [Rtk-users] FDK for planar ct
>
>         Hi,
>
>         Let me try to clarify what I mean by "source trajectory wrt
>         the object." In tomography, you need to determine the source
>         trajectory in the object coordinate system, we don't really
>         care about the source trajectory in the room coordinate
>         system. For example, rotating the source on a circular
>         trajectory or rotating the object makes no difference for the
>         reconstruction algorithm. That's why we call diagnostic
>         scanners "helical scanners".
>
>         So for trajectory 1, it seems that the source trajectory
>         (again, wrt to the object) is a circle but the object is
>         offset. This is somewhat similar to
>         https://doi.org/10.1109/TNS.2006.880977 except that the
>         detector is not tilted so FDK would be the only FBP algorithm
>         I could think of. But the situation is really not good, data
>         are missing and iterative reconstruction should give better
>         results.
>
>         Trajectory 2: what I said in my previous email is true, it's
>         useless I believe, all projections are similar up to a 2D
>         transform of the projection.
>
>         Simon
>
>         On Tue, Oct 10, 2017 at 8:07 PM, Robert Calliess
>         <robert.calliess at gmx.de <mailto:robert.calliess at gmx.de>> wrote:
>
>         Hello,
>
>         I try to clarify the both trajectories.
>
>         Trajectory 1:
>
>         No, i dont move the source on two circles. The xray source is
>         fixed. Only the object and the detector moves. Both move on a
>         circular path so that the iso-ray
>
>         always passes through the pcb centre and the detector centre.
>         There is one orthogonal view and the others are the ones
>         moving on the circular path.
>
>         (Object is not rotating around its own axis).
>
>         Trajectory 2:
>
>         Yes, the xray source lies in the rotation axis and only the
>         object rotates around its z-axis. Detector and xray source are
>         fixed and the detector is tilted.
>
>         It’s almost like this trajectory here
>         https://www.ikeda-shoponline.com/engctsoft/wp-content/uploads/2016/06/Oblique-View-CT1.jpg
>
>         except that the xray source lies on the rotation axis.
>
>         I hope this helps to understand the trajectories I have to
>         deal with.
>
>         Kind regards,
>
>         Robert
>
>         *Von:*simon.rit at gmail.com <mailto:simon.rit at gmail.com>
>         [mailto:simon.rit at gmail.com <mailto:simon.rit at gmail.com>] *Im
>         Auftrag von *Simon Rit
>         *Gese**ndet:*Dienstag, 10. Oktober 2017 19:06
>         *An:* Robert Calließ
>         *Cc:* Cyril Mory; rtk-users at public.kitware.com
>         <mailto:rtk-users at public.kitware.com>
>
>
>         *Betreff:* Re: [Rtk-users] FDK for planar ct
>
>         Hi,
>
>         It's still not clear to me but what is helpful is to think in
>         terms of source trajectory wrt the object.
>
>         Trajectory 1: if I understand, you move the source on two
>         circles plus one point. I don't know of a FBP algorithm to
>         reconstruct this, but there might be one. I would consider
>         iterative reconstruction first.
>
>         Trajectory 2: your trajectory is a point, the source does not
>         move with respect ot the object since it lies on the rotation
>         axis. So each projection contains exactly the same information
>         up to a simple 2D projection deformation. So it's hopeless to
>         reconstruct from one projection only.
>
>         To create the correct geometry, I would suggest using the
>         function AddProjection
>         <https://github.com/SimonRit/RTK/blob/master/code/rtkThreeDCircularProjectionGeometry.h#L92>
>         for which you provide the source and detector positions plus
>         the 3D coordinates of the two axes of the coordinate system of
>         the projection.
>
>         I hope this helps
>
>         Simon
>
>         On Tue, Oct 10, 2017 at 5:43 PM, "Robert Calließ"
>         <Robert.Calliess at gmx.de <mailto:Robert.Calliess at gmx.de>> wrote:
>
>         Hello,
>
>         thank you for the fast reply.
>
>         To answer your questions first.
>
>         In this case the abbrevation pcb stands for printed circuit board.
>
>         Next point is the trajectory we are currently handling with.
>
>         Please see the attached image "trajectory.png". There are two
>         schematics showing the side view and top view for trajectory
>         type 1
>
>         and a side-view for trajectory type 2.
>
>         For type 1:
>
>         The xray source is fixed. The pcb is clamped within a
>         transport, so the pcb and the detector are moveable with in
>         the xy plane.
>
>         As you can see at the image, the pcb moves along a circular
>         path but the pcb itself is not rotating. And let's assume that
>         the iso ray
>
>         always passes through the centre of the pcb and the centre of
>         the detector.
>
>         For type 2:
>
>         The xray source is fixed and the detector is tilted. The pcb
>         lies centred in the middle of a table. So that the pcb rotates
>         around its centre
>
>         around the z-axis.
>
>         I hope this makes clear what trajectory i'm dealing with.
>         Thank you.
>
>         Kind regards,
>
>         Robert C.
>
>         *Gesendet:* Dienstag, 10. Oktober 2017 um 15:31 Uhr
>         *Von:* "Cyril Mory" <cyril.mory at creatis.insa-lyon.fr
>         <mailto:cyril.mory at creatis.insa-lyon.fr>>
>         *An:* "Robert Calliess" <robert.calliess at gmx.de
>         <mailto:robert.calliess at gmx.de>>, rtk-users at public.kitware.com
>         <mailto:rtk-users at public.kitware.com>
>         *Betreff:* Re: [Rtk-users] FDK for planar ct
>
>         Dear Robert,
>
>         Your description of the trajectory is very obscure to me.
>         Maybe you have a very unusual X-ray system. Could you make the
>         following points clear :
>
>         - what is a PCB ?
>
>         - what is fixed/moving in your system (we need this
>         information for the object, the source and the detector), and
>         what kind of trajectories have the moving parts ?
>
>         - can you re-draw your sketch with just 2 or 3 positions
>         (ideally, on similar but separate drawings), each one with the
>         object, the source and the detector ?
>
>         If you do that, we should have a clear understanding of how
>         your acquisition goes, and be able to give you appropriate advice.
>
>         Best regards,
>
>         Cyril
>
>         On 10/10/2017 15:02, Robert Calliess wrote:
>
>             Hello rtk users,
>
>             I have question to the RTK FDK Filter. As far as I
>             understand from to the fourier slice theorem the object to
>             be reconstructed needs a circular trajectory and needs to
>             rotate its own centre.
>
>             Please have a look at the attached sketch. With this
>             planar trajectory (Object, a PCB, is moved on a circle
>             trajectpry  “in-plane”, PCB itself is not rotating) do I need
>
>             a special filtering if I want to use FDK for planar CT
>             with respect to the sketched trajectory ? I tried a
>             circular in-plane trajectory where the PCB is centred and
>             rotates
>
>             around its centre point. And with 100 projections I get
>             good results. But with the trajectory I described (sketch,
>             attached image) the results are not so good.
>
>             Because of the row-wise ramp filter It looks like there is
>             a directional dependency. My assumption is, and with
>             respect to fourier slice theorem, that the missing object
>
>             rotation (rotation around itself) causes there directional
>             effects.
>
>             So my questions to the experts are. Do I need to apply a
>             special filtering before backprojecting with FDK or is it
>             just the wrong
>
>             algorithm for this kind of trajectory ?
>
>             kind regards,
>
>             Robert C.
>
>             _______________________________________________
>
>             Rtk-users mailing list
>
>             Rtk-users at public.kitware.com
>             <mailto:Rtk-users at public.kitware.com>
>
>             http://public.kitware.com/mailman/listinfo/rtk-users
>
>
>         _______________________________________________
>         Rtk-users mailing list
>         Rtk-users at public.kitware.com <mailto:Rtk-users at public.kitware.com>
>         http://public.kitware.com/mailman/listinfo/rtk-users
>
>
>
> _______________________________________________
> Rtk-users mailing list
> Rtk-users at public.kitware.com
> http://public.kitware.com/mailman/listinfo/rtk-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/rtk-users/attachments/20171025/2f845852/attachment-0001.html>


More information about the Rtk-users mailing list