Re: [Rtk-users] More errors with 360 projections than 180 for SART
Hi, Thank you, this is probably the reason. Anaïs - Mail original - De: "Simon Rit" <simon@creatis.insa-lyon.fr> À: "Chao Wu" <wucha...@gmail.com> Cc: "anais capouillez" <anais.capouil...@student.uliege.be>, "rtk-users" <rtk-users@public.kitware.com> Envoyé: Jeudi 12 Avril 2018 07:50:54 Objet: Re: [Rtk-users] More errors with 360 projections than 180 for SART Hi Anais, The suggestion posted by Chao might be helpful to improve both results. My intuition is that this is caused by the fact that SART has no regularization and it is know that it will start to diverge from the solution after a few iterations. Setting the number of iteration is a way of avoiding divergence. Now, if you double the number of projections, I would say that one iteration goes two times faster to the solution. So I would compare a halved number of iterations when doubling the number of projections. I would suggest to use regularized conjugate gradient and to compare the results at convergence instead... I hope this helps, Simon On Wed, Apr 11, 2018 at 9:39 PM, Chao Wu <wucha...@gmail.com> wrote: > Hi Anais, > > I am travelling so I cannot review your images, but please check if the > following applies to you: > Open issue: https://github.com/SimonRit/RTK/issues/151 > Related thread: https://public.kitware.com/pipermail/rtk- > users/2018-February/010575.html > > Regards, > Chao > > 2018-04-11 15:37 GMT+02:00 <anais.capouil...@student.uliege.be>: > >> Hi, >> >> I'm reconstructing a homogeneous cylinder. My problem is that I have >> better results for 180 projections, than for 360 for SART (with default >> parameters), and the same geometry (except the number of projections). >> >> I don't understand why adding projections adds more errors, since it >> should be better (and for other algorithms this is indeed better with 360 >> projections). >> >> With 180 projections, the mean error on the cylinder is 1,6% , and the >> maximal error is 21,6%. While, with 360 projections, on the cylinder the >> mean error is 6,4%, and the maximal error is 169%. >> Even if I don't take into account voxels close to edges, there are more >> errors with 360 projections. >> >> Is there a reason why I obtain worse results with 360 projections? >> >> Since my files are too big, I added them here >> https://drive.google.com/drive/folders/1AYumjqNGUr8DZ0kxIbF7 >> Ab1xet7srk-z?usp=sharing. I also added a file with the commands I used. >> >> >> Thank you very much. >> >> Anaïs >> ___ >> Rtk-users mailing list >> Rtk-users@public.kitware.com >> https://public.kitware.com/mailman/listinfo/rtk-users >> > > > ___ > Rtk-users mailing list > Rtk-users@public.kitware.com > https://public.kitware.com/mailman/listinfo/rtk-users > > ___ Rtk-users mailing list Rtk-users@public.kitware.com https://public.kitware.com/mailman/listinfo/rtk-users
[Rtk-users] More errors with 360 projections than 180 for SART
Hi, I'm reconstructing a homogeneous cylinder. My problem is that I have better results for 180 projections, than for 360 for SART (with default parameters), and the same geometry (except the number of projections). I don't understand why adding projections adds more errors, since it should be better (and for other algorithms this is indeed better with 360 projections). With 180 projections, the mean error on the cylinder is 1,6% , and the maximal error is 21,6%. While, with 360 projections, on the cylinder the mean error is 6,4%, and the maximal error is 169%. Even if I don't take into account voxels close to edges, there are more errors with 360 projections. Is there a reason why I obtain worse results with 360 projections? Since my files are too big, I added them here https://drive.google.com/drive/folders/1AYumjqNGUr8DZ0kxIbF7Ab1xet7srk-z?usp=sharing. I also added a file with the commands I used. Thank you very much. Anaïs ___ Rtk-users mailing list Rtk-users@public.kitware.com https://public.kitware.com/mailman/listinfo/rtk-users
Re: [Rtk-users] Reducing the pattern of errors
Thank you very much. This is probably due to the resolution I use then. I cannot use a better resolution for now because there is not enough memory when I use Cuda (but I will work with a better graphic card soon). Best regards. Anaïs - Mail original - De: "Simon Rit" <simon@creatis.insa-lyon.fr> À: "anais capouillez" <anais.capouil...@student.uliege.be> Cc: rtk-users@public.kitware.com Envoyé: Mercredi 14 Février 2018 16:30:49 Objet: Re: [Rtk-users] Reducing the pattern of errors Hi, These are numerical errors. How did you create the projections? It seems that you have used the voxelized phantom to simulate the projections. This enhances strongly these artefacts. This is why most people compute simulated projections analytically with simple analytical shapes (as the Shepp Logan phantom). If you share your projections meta information, I can illustrate how to do the simulation differently. If you already used analytical simulations, a way to improve the quality is to increase the sampling (use more and finer pixels in the projections). If you want to reduce the artefacts from these projections without modifying the projections sampling, you need to remove some high frequencies with a proper windowing. You can try --hann 1 for example on the rtkfdk command line. I hope this helps, Simon On Wed, Feb 14, 2018 at 1:47 PM, <anais.capouil...@student.uliege.be> wrote: > Hi, > > When I use FDK to reconstruct my volume, there are some parts inside the > volume where the error is up to 10%, even in the middle of my volume. When > I compute the difference between the phantom and the reconstructed object I > can see some patterns of small lines where the error is way bigger than in > the rest of the volume. > I need to reconstruct my volume with at least one big area with no error > bigger than few percents of error for each voxel of the area. The patterns > of errors occur too often to select an area sufficiently big. > Unfortunately, I cannot change the number of projections I use because > this is imposed to me. > > Therefore, I want to know if it is possible to obtain better results with > FDK or if I have to use an iterative algorithm. > > > > For the parameters of the geometry, I used 180 projections, sdd=978.5, and > sid=478.5. For the projections I used a spacing of 0.8 and a dimension of > 1024. And for the reconstruction, I used a spacing of 0.5, and a dimension > of 204*404*204. > > I joined two screenshots of the absolute difference between the phantom > and the reconstruction (one zoomed on some of the small lines of error, and > another one not zoomed). > > > If you want the images of my phantom, the reconstruction, and the absolute > difference between the two (with the actual values and with relative > values), I uploaded them here: https://drive.google.com/drive/folders/ > 194k2CDomeLlmVxybTllSKWYhpCZyPvVx?usp=sharing > > > Thank you. > > Anaïs > ___ > Rtk-users mailing list > Rtk-users@public.kitware.com > https://public.kitware.com/mailman/listinfo/rtk-users > > ___ Rtk-users mailing list Rtk-users@public.kitware.com https://public.kitware.com/mailman/listinfo/rtk-users
Re: [Rtk-users] Problems with --like and --dimension parameters of rtkforwardprojections and rtkfdk
Hi Simon, I have just redone an analysis and everything went through with no issue. My advisor and I probably messed up the values of spacing and/or dimension when we used the rtkforwardprojections. Thank you for the quick reply and sorry for wasting a little bit of your time. Best regards. Anaïs - Mail original - De: "Simon Rit" <simon@creatis.insa-lyon.fr> À: "anais capouillez" <anais.capouil...@student.uliege.be> Cc: rtk-users@public.kitware.com Envoyé: Lundi 29 Janvier 2018 12:33:20 Objet: Re: [Rtk-users] Problems with --like and --dimension parameters of rtkforwardprojections and rtkfdk Hi Anais, It's a bit hard to check what's going on without the command lines. Can you try to send the command lines that give you some problems, e.g. based on the forwardprojection example? Thanks, Simon On Mon, Jan 29, 2018 at 12:03 PM, <anais.capouil...@student.uliege.be> wrote: > Hi, > > My name is Anaïs, I’m a student in computer sciences, and I'm currently > doing my internship, which consists partly to do a tomographic > reconstruction. > > I use rtk to reconstruct my model, but I have problems when I use > rtkforwardprojections and rtkfdk with –like or –dimension. > Indeed, if I use them with the default parameters, it works correctly, but > when I specify –like or certain values of –dimension the reconstruction > isn’t correct. > For instance if I use the example file muCT.mha as my model, it works > perfectly with the default parameters, but if I choose –like=muCT.mha, in > my reconstruction file there are mostly black voxels. > I also saw that the example in the scripts for forward projections ( > http://wiki.openrtk.org/index.php/RTK/Scripts/ForwardProjection) works > when the spacing, the origin, and the dimension specified to rtkfdk are the > same than the model. > > I wanted to know why it doesn’t works when I use –like. > Is there a relation between the geometry and dimensions and that’s why I > have problems (for instance due to the source being too close to the > object)? > > > Thank you very much. > > Anaïs > ___ > Rtk-users mailing list > Rtk-users@public.kitware.com > https://public.kitware.com/mailman/listinfo/rtk-users > ___ Rtk-users mailing list Rtk-users@public.kitware.com https://public.kitware.com/mailman/listinfo/rtk-users
[Rtk-users] Problems with --like and --dimension parameters of rtkforwardprojections and rtkfdk
Hi, My name is Anaïs, I’m a student in computer sciences, and I'm currently doing my internship, which consists partly to do a tomographic reconstruction. I use rtk to reconstruct my model, but I have problems when I use rtkforwardprojections and rtkfdk with –like or –dimension. Indeed, if I use them with the default parameters, it works correctly, but when I specify –like or certain values of –dimension the reconstruction isn’t correct. For instance if I use the example file muCT.mha as my model, it works perfectly with the default parameters, but if I choose –like=muCT.mha, in my reconstruction file there are mostly black voxels. I also saw that the example in the scripts for forward projections (http://wiki.openrtk.org/index.php/RTK/Scripts/ForwardProjection) works when the spacing, the origin, and the dimension specified to rtkfdk are the same than the model. I wanted to know why it doesn’t works when I use –like. Is there a relation between the geometry and dimensions and that’s why I have problems (for instance due to the source being too close to the object)? Thank you very much. Anaïs ___ Rtk-users mailing list Rtk-users@public.kitware.com https://public.kitware.com/mailman/listinfo/rtk-users