[Rtk-users] Sophia beads reconstruction
Simon Rit
simon.rit at creatis.insa-lyon.fr
Fri Feb 5 09:32:08 EST 2016
Hi MG,
Sorry, I realized I said something stupid. Using tif or mha won't make
any difference because there is a branching on the pixel type in the
projectionsreader
<http://www.openrtk.org/Doxygen/classrtk_1_1ProjectionsReader.html>
(search for ushort in the graph of the doxygen doc
<http://www.openrtk.org/Doxygen/classrtk_1_1ProjectionsReader.html>) and
unsigned short are handled differently.
The enclosed pyhon code will reproduce the same as my previous command
line application. The commented code in it is what I initially used to
read projections without any processing, i.e., assuming line integrals.
I still think there is a problem in the geometry but someone (you?) has
to spend sometime trying to understand what are the parameters from the
Matlab code.
Keep us posted and good luck,
Simon
On 28/01/2016 22:27, MG Vallet wrote:
> Louie,
> I played with sdd and sid, following your advise. But they have few
> effect, clearly less effect than proj_iso_x. And none of these
> parameters can significantly improve my result.
>
> Simon,
> I carefully checked every input data and I cant see what I did wrong.
> I want first to reproduce your output. You said earlier it's easy to
> convert a series of tiff image into a mha file using SimpleRTK. Do you
> mean by writing a short program, or is it possible from the command
> line ?
>
> MG
>
> 2016-01-26 17:17 GMT-05:00 louie L <ghostcz at hotmail.com
> <mailto:ghostcz at hotmail.com>>:
>
> Hi MG,
>
> Sometimes, the progressive deformation from the center of rotation
> to the edge of the image is caused by wrong sdd/sid ratio. Such as
> change in the geometry will lead to distortion of the image and
> wrong magnification.
>
> Maybe you can play with that parameter and report your result.
>
> Cheers,
>
> Louie
>
> Sent from Mail <http://go.microsoft.com/fwlink/?LinkId=550986> for
> Windows 10
>
>
> *From: *Simon Rit <mailto:simon.rit at creatis.insa-lyon.fr>
> *Sent: *Tuesday, January 26, 2016 13:24
> *To: *MG Vallet <mailto:mgvallet.pro at gmail.com>
> *Cc: *rtk-users at public.kitware.com
> <mailto:rtk-users at public.kitware.com>
> *Subject: *Re: [Rtk-users] Sophia beads reconstruction
>
> Hi,
>
> I think it's likely a problem with parameters. The best would be
> to obtain them from the Matlab code they provide.
>
> Good luck,
>
> Simon
>
> On Mon, Jan 25, 2016 at 5:48 PM, MG Vallet <mgvallet.pro at gmail.com
> <mailto:mgvallet.pro at gmail.com>> wrote:
>
> Hi Simon,
>
> this was a point I missed. I had to do the log mapping.
>
> I must say, I do not use RTK from shell. I just plugged the
> objects
>
> FDKConeBeamReconstructionFilter
> and ThreeDCircularProjectionGeometry from RTK into a
> visualisation software. That's the reason I'm not absolutely
> sure about the length parameter setting.
>
> Anyway, I'm now pretty close to your result. I believe the
> only difference is the progressive deformation of beads with
> the distance to center.
>
> What's your opinion : is there another step I might have
> missed, or is it more likely a problem with parameters ?
>
> MG
>
> 2016-01-21 14:22 GMT-05:00 Simon Rit
> <simon.rit at creatis.insa-lyon.fr
> <mailto:simon.rit at creatis.insa-lyon.fr>>:
>
> Not exactly. rtk::FDKConeBeamProjectionFilter assumes that
> you pass the line integral (ln(I0/I)). But rtkfdk uses the
> ProjectionsReader to read the projections and converts
> them depending on the type: for mha or mhd, it just reads
> it and nothing else. For tif, it assumes that I is passed,
> try to guess I0 (or gets it from the cmd line) and
> computes ln(I0/I).
> So from each tif projection, instead of letting them go
> through the ProjectionsReader in RTK, directly use the
> ImageReader, stack them and pass it to the reconstruction
> algorithm. In other words, if you use SimpleRTK, use
> ImageSeriesReader instead of ProjectionsReader.
> Regarding the geometry, it would be nice to be confident
> in the geometry parameters, maybe from their Matlab code?
> Simon
>
> On 21/01/2016 20:02, MG Vallet wrote:
>
> Hi Simon and Cyril,
>
> I really appreciate your insight.
>
> By examining the parameters of the acquisition, I saw
> that the center must effectively be offset (4.8 mm in
> both directions). For the attached image, I also
> increases sdd and sid by a factor 1000, just in case I
> would be wrong in units. It reduces the first default
> I pointed out in my first message. Also note that my
> input dataset was SophiaBeads_512_averaged.
>
> Regarding the pre-processing mentioned by Simon, I
> need more help. Is it as simple as the FDK filter
> assumes that the input is intensity, but the dataset
> provides -ln(I/I0) ?
>
> Regards,
>
> mg
>
> 2016-01-20 3:33 GMT-05:00 Simon Rit
> <simon.rit at creatis.insa-lyon.fr
> <mailto:simon.rit at creatis.insa-lyon.fr>>:
>
> Hi,
>
> I quickly checked SophiaBeads_64_averaged.zip.
> First, it seems that what is provided is directly
> the line integral. RTK will process the tif images
> as if it were a photon count so I converted the
> tif projections to mha to avoid this. For the
> conversion, I used another piece of software but
> you can easily do the same thing with SimpleRTK.
>
> Second, I used the enclosed script which gives me
> a much better result. I have however a geometry
> problem. I think they provide a Matlab piece of
> code to get the center of rotation. If you can run
> it and find out what is this center, then you can
> use it to configure the geometry better with the
> options --proj_iso_x and --proj_iso_y.
>
> Good luck,
>
> Simon
>
> On Tue, Jan 19, 2016 at 8:02 PM, MG Vallet
> <mgvallet.pro at gmail.com
> <mailto:mgvallet.pro at gmail.com>> wrote:
>
> Thanks Cyril!
>
> going to carefully check my inputs and I give
> you some news.
>
> mgv
>
> 2016-01-19 10:33 GMT-05:00 Cyril Mory
> <cyril.mory at uclouvain.be
> <mailto:cyril.mory at uclouvain.be>>:
>
> Hi,
>
> It looks like the geometry is wrong
> indeed, because I have never seen such
> artifacts.
>
> Hope that helps,
> Cyril
>
> Le 19/01/2016 16:19, MG Vallet a écrit :
>
> Dear rtk-users,
>
> I applied the
> FDKConeBeamReconstructionFilter to a
> data set from
> https://zenodo.org/record/16539.
>
> This is a plastic tube filled with
> glass beads that has been scanned.
>
> Attached is a view of the
> reconstructed central slice. It is
> found that:
>
> 1. the beads are quickly deformed
> outside the center
> 2. each one has a halo, or a shadow
> (dont know the proper term)
>
> My question is if these are simply
> expected artifacts. I am not 100% sure
> I properly set the geometry parameters.
>
> Regards,
>
> mgv
>
> _______________________________________________
>
> 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
>
>
>
>
>
>
> L'absence de virus dans ce courrier
> électronique a été vérifiée par le
> logiciel antivirus Avast.
> www.avast.com
> <https://www.avast.com/antivirus>
>
>
> _______________________________________________
> 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
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/rtk-users/attachments/20160205/3cfb5166/attachment-0009.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: fdk.py
Type: text/x-python-script
Size: 1518 bytes
Desc: not available
URL: <http://public.kitware.com/pipermail/rtk-users/attachments/20160205/3cfb5166/attachment-0009.bin>
More information about the Rtk-users
mailing list