From simon.rit at creatis.insa-lyon.fr Fri Mar 1 02:52:08 2013 From: simon.rit at creatis.insa-lyon.fr (Simon Rit) Date: Fri, 1 Mar 2013 08:52:08 +0100 Subject: [Rtk-users] trouble projecting phantom In-Reply-To: <1361943948.94604.YahooMailClassic@web163804.mail.gq1.yahoo.com> References: <1361943948.94604.YahooMailClassic@web163804.mail.gq1.yahoo.com> Message-ID: The size of your image is determined by origin, number of pixels (dimension in command line tools or size in itk) and spacing. From this information, you can check if the source to isocenter (point 0,0,0) distance puts the source in or out of the volume. I will try to correct this bug asap possible, Simon On Wed, Feb 27, 2013 at 6:45 AM, M Miller wrote: > Scaling the phantom worked perfectly. > You mentioned that a reconstruction will need to specify the source > outside of the volume. How can I specify that? What parameters would you > pass to rtkfdk.exe to reconstruct the phantom? > > Thanks > Micah > > --- On *Tue, 2/26/13, Simon Rit * wrote: > > > From: Simon Rit > Subject: Re: [Rtk-users] trouble projecting phantom > To: "M Miller" > Cc: rtk-users at openrtk.org > Date: Tuesday, February 26, 2013, 12:36 AM > > > Hi Micah, > I have given it a try and I think I obtain the expected result if you use > the SheppLogan file that we provide here > http://midas3.kitware.com/midas/download?items=27326 > However, the result might surprise you because your source is in the > phantom (sid=65.93 vs largest ellipsoid which is A=88.32 B=115.2 C=117.76). > You can scale the shepp logan phantom using the following command > rtkprojectshepploganphantom -g geo.xml -o proj.mha --spacing 1 --dimension > 1024 --phantomscale 10 > You will then obtain more meaningful results, i.e., the full phantom will > be in the beam. > Note by the way that if you use rtkfdk afterwards, you will have to be > careful to have the source outside the volume as well. I have recently > noticed that there was a bug when the source was in the volume, I will fix > it as soon as possible. > Simon > > On Tue, Feb 26, 2013 at 2:23 AM, M Miller > > wrote: > > I've been unable to project a phantom, or anything else, at high > magnification. Here's the steps I've think should work, but don't. > > "rtksimulatedgeometry.exe --output=geo.xml --nproj=1600 --arc=360 > --sdd=870.86 --sid=65.93" > > These are the values from my actual CT system. Next I try using > rtkprojectgeometricphantom to create some projections. > > "rtkprojectgeometricphantom -g geo.xml -o projs\projections1024.mha > --spacing 1 --dimension 1024 --phantomfile > SheppLogan.txt" > > Inspecting the output projections shows they're mostly unrecognizable. > Repeating these steps with a higher SID does produce the expected phantom > projections. I've had no luck using real data, so I thought I'd try a > simple phantom, but it's turned out to be not so simple. > > Is RTK limited somehow with respect to magnification, or are my > expectations unrealsitic? Am I missing a command-line switch? > > Thanks > Micah > > _______________________________________________ > Rtk-users mailing list > Rtk-users at openrtk.org > http://public.kitware.com/cgi-bin/mailman/listinfo/rtk-users > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From croow at yahoo.com Tue Mar 5 20:07:53 2013 From: croow at yahoo.com (M Miller) Date: Tue, 5 Mar 2013 17:07:53 -0800 (PST) Subject: [Rtk-users] Parallel beam? Message-ID: <1362532073.88670.YahooMailClassic@web163801.mail.gq1.yahoo.com> Can rtk reconstruct parallel beam data (circular trajectory)? If so, what would be the best way to create the geometry file? Would rtkfdk.exe and rtksart.exe be appropriate? ? Thanks -------------- next part -------------- An HTML attachment was scrubbed... URL: From simon.rit at creatis.insa-lyon.fr Wed Mar 6 01:28:30 2013 From: simon.rit at creatis.insa-lyon.fr (Simon Rit) Date: Wed, 6 Mar 2013 07:28:30 +0100 Subject: [Rtk-users] Parallel beam? In-Reply-To: <1362532073.88670.YahooMailClassic@web163801.mail.gq1.yahoo.com> References: <1362532073.88670.YahooMailClassic@web163801.mail.gq1.yahoo.com> Message-ID: Hi, We have started working on it but many things are not yet ready. Only rtkfdk should work by setting SourceToDetectorDistance to 0. The rest (forward projection, sart, shepp logan simulations, etc.) has not been adapted yet. Simon On Wed, Mar 6, 2013 at 2:07 AM, M Miller wrote: > Can rtk reconstruct parallel beam data (circular trajectory)? > If so, what would be the best way to create the geometry file? > Would rtkfdk.exe and rtksart.exe be appropriate? > > Thanks > > _______________________________________________ > Rtk-users mailing list > Rtk-users at openrtk.org > http://public.kitware.com/cgi-bin/mailman/listinfo/rtk-users > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From simon.rit at creatis.insa-lyon.fr Thu Mar 7 21:47:48 2013 From: simon.rit at creatis.insa-lyon.fr (Simon Rit) Date: Fri, 8 Mar 2013 03:47:48 +0100 Subject: [Rtk-users] trouble projecting phantom In-Reply-To: References: <1361943948.94604.YahooMailClassic@web163804.mail.gq1.yahoo.com> Message-ID: Hi, I have pushed a fix so it should work now, but be aware that putting your source in the volume has no sense with FDK because the weighted backprojection in the algorithm consists in dividing by the distance to the source. So it implies a division by 0. Simon On Fri, Mar 1, 2013 at 8:52 AM, Simon Rit wrote: > The size of your image is determined by origin, number of pixels > (dimension in command line tools or size in itk) and spacing. From this > information, you can check if the source to isocenter (point 0,0,0) > distance puts the source in or out of the volume. > I will try to correct this bug asap possible, > Simon > > > On Wed, Feb 27, 2013 at 6:45 AM, M Miller wrote: > >> Scaling the phantom worked perfectly. >> You mentioned that a reconstruction will need to specify the source >> outside of the volume. How can I specify that? What parameters would you >> pass to rtkfdk.exe to reconstruct the phantom? >> >> Thanks >> Micah >> >> --- On *Tue, 2/26/13, Simon Rit * wrote: >> >> >> From: Simon Rit >> Subject: Re: [Rtk-users] trouble projecting phantom >> To: "M Miller" >> Cc: rtk-users at openrtk.org >> Date: Tuesday, February 26, 2013, 12:36 AM >> >> >> Hi Micah, >> I have given it a try and I think I obtain the expected result if you use >> the SheppLogan file that we provide here >> http://midas3.kitware.com/midas/download?items=27326 >> However, the result might surprise you because your source is in the >> phantom (sid=65.93 vs largest ellipsoid which is A=88.32 B=115.2 C=117.76). >> You can scale the shepp logan phantom using the following command >> rtkprojectshepploganphantom -g geo.xml -o proj.mha --spacing 1 >> --dimension 1024 --phantomscale 10 >> You will then obtain more meaningful results, i.e., the full phantom will >> be in the beam. >> Note by the way that if you use rtkfdk afterwards, you will have to be >> careful to have the source outside the volume as well. I have recently >> noticed that there was a bug when the source was in the volume, I will fix >> it as soon as possible. >> Simon >> >> On Tue, Feb 26, 2013 at 2:23 AM, M Miller >> > wrote: >> >> I've been unable to project a phantom, or anything else, at high >> magnification. Here's the steps I've think should work, but don't. >> >> "rtksimulatedgeometry.exe --output=geo.xml --nproj=1600 --arc=360 >> --sdd=870.86 --sid=65.93" >> >> These are the values from my actual CT system. Next I try using >> rtkprojectgeometricphantom to create some projections. >> >> "rtkprojectgeometricphantom -g geo.xml -o projs\projections1024.mha >> --spacing 1 --dimension 1024 --phantomfile >> SheppLogan.txt" >> >> Inspecting the output projections shows they're mostly unrecognizable. >> Repeating these steps with a higher SID does produce the expected phantom >> projections. I've had no luck using real data, so I thought I'd try a >> simple phantom, but it's turned out to be not so simple. >> >> Is RTK limited somehow with respect to magnification, or are my >> expectations unrealsitic? Am I missing a command-line switch? >> >> Thanks >> Micah >> >> _______________________________________________ >> Rtk-users mailing list >> Rtk-users at openrtk.org >> http://public.kitware.com/cgi-bin/mailman/listinfo/rtk-users >> >> >> > -------------- next part -------------- An HTML attachment was scrubbed... URL: From hsieandy at gmail.com Mon Mar 11 02:32:15 2013 From: hsieandy at gmail.com (Andy Shieh) Date: Mon, 11 Mar 2013 17:32:15 +1100 Subject: [Rtk-users] About SART Message-ID: Hi Simon, I have a question regarding practical implementation of SART. As far as I know, one needs to find the ray-sum coefficient w_ij by bilinear interpolations to simulate the forward projection process. Then, the backprojection process is done in an algebraic manner using the w_ij we found according to the SART formula. However, according to a few open source codes I've found, as well as rtksart, I realize in many cases people simply use a traditional backprojection (with or without ramp filter) for the backprojection step. Is there any reference, or justification, that this is equivalent, or approximately equivalent, to the actual SART? Thanks! Not sure if I've described the question clear enough. Please let me know if it doesn't make much sense and I'll try to elaborate a bit more. Thanks :) Cheers, Andy -------------- next part -------------- An HTML attachment was scrubbed... URL: From simon.rit at creatis.insa-lyon.fr Mon Mar 11 02:56:03 2013 From: simon.rit at creatis.insa-lyon.fr (Simon Rit) Date: Mon, 11 Mar 2013 07:56:03 +0100 Subject: [Rtk-users] About SART In-Reply-To: References: Message-ID: Hi Andy, If I understand you correctly, yes there is: http://www.ucair.med.utah.edu/FacultyZeng/zeng-papers/Unmatched2000.pdf But in rtksart, you can change the backprojector to a matched one with the option --bp Joseph. Note that rtksart has not been tested in depth so let us know if you encounter problems, Simon On Mon, Mar 11, 2013 at 7:32 AM, Andy Shieh wrote: > Hi Simon, > > I have a question regarding practical implementation of SART. As far as I > know, one needs to find the ray-sum coefficient w_ij by bilinear > interpolations to simulate the forward projection process. Then, the > backprojection process is done in an algebraic manner using the w_ij we > found according to the SART formula. However, according to a few open source > codes I've found, as well as rtksart, I realize in many cases people simply > use a traditional backprojection (with or without ramp filter) for the > backprojection step. Is there any reference, or justification, that this is > equivalent, or approximately equivalent, to the actual SART? Thanks! > > Not sure if I've described the question clear enough. Please let me know if > it doesn't make much sense and I'll try to elaborate a bit more. Thanks :) > > Cheers, > Andy > > _______________________________________________ > Rtk-users mailing list > Rtk-users at openrtk.org > http://public.kitware.com/cgi-bin/mailman/listinfo/rtk-users > From seryongkang at naver.com Mon Mar 11 04:28:52 2013 From: seryongkang at naver.com (=?UTF-8?B?6rCV7IS466Oh?=) Date: Mon, 11 Mar 2013 17:28:52 +0900 (KST) Subject: [Rtk-users] =?utf-8?q?__About_SART_with_CUDA?= Message-ID: Hello , simon. I used RTK for SART with CUDA, GPU. I have used this code with geforce GTX 560, CUDA SDK 4.1,and it was working greatly. But, recently I changed my graphic card to geforce GTX 680, and then the RTK SART code have been not working. The error message is itk::ERROR:: CUDA ERROR: invalid device function. How can I fix my problem. Please help me. Thanks. -------------- next part -------------- An HTML attachment was scrubbed... URL: From simon.rit at creatis.insa-lyon.fr Mon Mar 11 06:03:16 2013 From: simon.rit at creatis.insa-lyon.fr (Simon Rit) Date: Mon, 11 Mar 2013 11:03:16 +0100 Subject: [Rtk-users] About SART with CUDA In-Reply-To: References: Message-ID: Hi, This can happen when the requested ressources are not in accordance with the GPU characteristics, e.g., wrong block sizes, texture size, etc. Try to locate the line giving you this error (it's given in the exception error message) and modify the hard coded parameters. We can help if you provide the complete error message. Simon On Mon, Mar 11, 2013 at 9:28 AM, ??? wrote: > Hello , simon. > > > > I used RTK for SART with CUDA, GPU. > > > > I have used this code with geforce GTX 560, CUDA SDK 4.1,and it was > working greatly. > > > > But, recently I changed my graphic card to geforce GTX 680, and then the > RTK SART code have been not working. > > > > The error message is itk::ERROR:: CUDA ERROR: invalid device function. > > > > How can I fix my problem. Please help me. > > > > Thanks. > > _______________________________________________ > Rtk-users mailing list > Rtk-users at openrtk.org > http://public.kitware.com/cgi-bin/mailman/listinfo/rtk-users > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From simon.rit at creatis.insa-lyon.fr Mon Mar 11 08:13:31 2013 From: simon.rit at creatis.insa-lyon.fr (Simon Rit) Date: Mon, 11 Mar 2013 13:13:31 +0100 Subject: [Rtk-users] About SART with CUDA In-Reply-To: References: <911ac3fb64426d877dccaebe325fd6d2@tweb03.nm.nhnsystem.com> Message-ID: Please continue using the mailing list. It's hard too say, maybe the volume is too big or too small for your GPU. Have you tried doing "make test" to see if the problem is related to this particular reconstruction or if this is a more general problem? What is the size of the projection images and the volume that you use? Simon On Mon, Mar 11, 2013 at 11:52 AM, ??? wrote: > Hi. > > > > This is the line giving this error. > > > > : In rtkCudaForwardProjectionImageFilter.cu line : 251 >> error > message is invalid texture reference > > > > // Set texture parameters > > tex_vol.normalized = false; // access with > normalized texture coordinates > > tex_vol.filterMode = cudaFilterModeLinear; // linear interpolation > > tex_vol.addressMode[0] = cudaAddressModeClamp; // clamp texture > coordinates > > tex_vol.addressMode[1] = cudaAddressModeClamp; > > // Bind 3D array to 3D texture > > cudaBindTextureToArray(tex_vol, (cudaArray*)dev_vol, channelDesc); > > CUDA_CHECK_ERROR; --------------->>> this is line 251 > > > > > > and : > > > > In rtkCudaBackProjectionImageFilter,.cu line : 251 >> error message is > invalid device function > > > > kernel <<< dimGrid, dimBlock >>> ( dev_vol, > > make_int3(vol_dim[0], > vol_dim[1], vol_dim[2]), > > blocksInY, > 1.0f/(float)blocksInY ); > > } > > CUDA_CHECK_ERROR; ------------>>> this is line 251 > > > > > > I can't understand about this error..... > > > > Please help me, simon. > > > > > > Thanks. > > -seryong > -------------- next part -------------- An HTML attachment was scrubbed... URL: From hsieandy at gmail.com Mon Mar 11 19:29:05 2013 From: hsieandy at gmail.com (Andy Shieh) Date: Tue, 12 Mar 2013 10:29:05 +1100 Subject: [Rtk-users] About SART In-Reply-To: References:

Message-ID: Hi Simon, Yes, this is exactly what I'm looking for! Thanks so much for your help! Cheers, Andy 2013/3/11 Simon Rit > Hi Andy, > If I understand you correctly, yes there is: > http://www.ucair.med.utah.edu/FacultyZeng/zeng-papers/Unmatched2000.pdf > But in rtksart, you can change the backprojector to a matched one with > the option --bp Joseph. Note that rtksart has not been tested in depth > so let us know if you encounter problems, > Simon > > On Mon, Mar 11, 2013 at 7:32 AM, Andy Shieh wrote: > > Hi Simon, > > > > I have a question regarding practical implementation of SART. As far as I > > know, one needs to find the ray-sum coefficient w_ij by bilinear > > interpolations to simulate the forward projection process. Then, the > > backprojection process is done in an algebraic manner using the w_ij we > > found according to the SART formula. However, according to a few open > source > > codes I've found, as well as rtksart, I realize in many cases people > simply > > use a traditional backprojection (with or without ramp filter) for the > > backprojection step. Is there any reference, or justification, that this > is > > equivalent, or approximately equivalent, to the actual SART? Thanks! > > > > Not sure if I've described the question clear enough. Please let me know > if > > it doesn't make much sense and I'll try to elaborate a bit more. Thanks > :) > > > > Cheers, > > Andy > > > > _______________________________________________ > > Rtk-users mailing list > > Rtk-users at openrtk.org > > http://public.kitware.com/cgi-bin/mailman/listinfo/rtk-users > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From simon.rit at creatis.insa-lyon.fr Fri Mar 1 02:52:08 2013 From: simon.rit at creatis.insa-lyon.fr (Simon Rit) Date: Fri, 1 Mar 2013 08:52:08 +0100 Subject: [Rtk-users] trouble projecting phantom In-Reply-To: <1361943948.94604.YahooMailClassic@web163804.mail.gq1.yahoo.com> References: <1361943948.94604.YahooMailClassic@web163804.mail.gq1.yahoo.com> Message-ID: The size of your image is determined by origin, number of pixels (dimension in command line tools or size in itk) and spacing. From this information, you can check if the source to isocenter (point 0,0,0) distance puts the source in or out of the volume. I will try to correct this bug asap possible, Simon On Wed, Feb 27, 2013 at 6:45 AM, M Miller wrote: > Scaling the phantom worked perfectly. > You mentioned that a reconstruction will need to specify the source > outside of the volume. How can I specify that? What parameters would you > pass to rtkfdk.exe to reconstruct the phantom? > > Thanks > Micah > > --- On *Tue, 2/26/13, Simon Rit * wrote: > > > From: Simon Rit > Subject: Re: [Rtk-users] trouble projecting phantom > To: "M Miller" > Cc: rtk-users at openrtk.org > Date: Tuesday, February 26, 2013, 12:36 AM > > > Hi Micah, > I have given it a try and I think I obtain the expected result if you use > the SheppLogan file that we provide here > http://midas3.kitware.com/midas/download?items=27326 > However, the result might surprise you because your source is in the > phantom (sid=65.93 vs largest ellipsoid which is A=88.32 B=115.2 C=117.76). > You can scale the shepp logan phantom using the following command > rtkprojectshepploganphantom -g geo.xml -o proj.mha --spacing 1 --dimension > 1024 --phantomscale 10 > You will then obtain more meaningful results, i.e., the full phantom will > be in the beam. > Note by the way that if you use rtkfdk afterwards, you will have to be > careful to have the source outside the volume as well. I have recently > noticed that there was a bug when the source was in the volume, I will fix > it as soon as possible. > Simon > > On Tue, Feb 26, 2013 at 2:23 AM, M Miller > > wrote: > > I've been unable to project a phantom, or anything else, at high > magnification. Here's the steps I've think should work, but don't. > > "rtksimulatedgeometry.exe --output=geo.xml --nproj=1600 --arc=360 > --sdd=870.86 --sid=65.93" > > These are the values from my actual CT system. Next I try using > rtkprojectgeometricphantom to create some projections. > > "rtkprojectgeometricphantom -g geo.xml -o projs\projections1024.mha > --spacing 1 --dimension 1024 --phantomfile > SheppLogan.txt" > > Inspecting the output projections shows they're mostly unrecognizable. > Repeating these steps with a higher SID does produce the expected phantom > projections. I've had no luck using real data, so I thought I'd try a > simple phantom, but it's turned out to be not so simple. > > Is RTK limited somehow with respect to magnification, or are my > expectations unrealsitic? Am I missing a command-line switch? > > Thanks > Micah > > _______________________________________________ > Rtk-users mailing list > Rtk-users at openrtk.org > http://public.kitware.com/cgi-bin/mailman/listinfo/rtk-users > > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From croow at yahoo.com Tue Mar 5 20:07:53 2013 From: croow at yahoo.com (M Miller) Date: Tue, 5 Mar 2013 17:07:53 -0800 (PST) Subject: [Rtk-users] Parallel beam? Message-ID: <1362532073.88670.YahooMailClassic@web163801.mail.gq1.yahoo.com> Can rtk reconstruct parallel beam data (circular trajectory)? If so, what would be the best way to create the geometry file? Would rtkfdk.exe and rtksart.exe be appropriate? ? Thanks -------------- next part -------------- An HTML attachment was scrubbed... URL: From simon.rit at creatis.insa-lyon.fr Wed Mar 6 01:28:30 2013 From: simon.rit at creatis.insa-lyon.fr (Simon Rit) Date: Wed, 6 Mar 2013 07:28:30 +0100 Subject: [Rtk-users] Parallel beam? In-Reply-To: <1362532073.88670.YahooMailClassic@web163801.mail.gq1.yahoo.com> References: <1362532073.88670.YahooMailClassic@web163801.mail.gq1.yahoo.com> Message-ID: Hi, We have started working on it but many things are not yet ready. Only rtkfdk should work by setting SourceToDetectorDistance to 0. The rest (forward projection, sart, shepp logan simulations, etc.) has not been adapted yet. Simon On Wed, Mar 6, 2013 at 2:07 AM, M Miller wrote: > Can rtk reconstruct parallel beam data (circular trajectory)? > If so, what would be the best way to create the geometry file? > Would rtkfdk.exe and rtksart.exe be appropriate? > > Thanks > > _______________________________________________ > Rtk-users mailing list > Rtk-users at openrtk.org > http://public.kitware.com/cgi-bin/mailman/listinfo/rtk-users > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From simon.rit at creatis.insa-lyon.fr Thu Mar 7 21:47:48 2013 From: simon.rit at creatis.insa-lyon.fr (Simon Rit) Date: Fri, 8 Mar 2013 03:47:48 +0100 Subject: [Rtk-users] trouble projecting phantom In-Reply-To: References: <1361943948.94604.YahooMailClassic@web163804.mail.gq1.yahoo.com> Message-ID: Hi, I have pushed a fix so it should work now, but be aware that putting your source in the volume has no sense with FDK because the weighted backprojection in the algorithm consists in dividing by the distance to the source. So it implies a division by 0. Simon On Fri, Mar 1, 2013 at 8:52 AM, Simon Rit wrote: > The size of your image is determined by origin, number of pixels > (dimension in command line tools or size in itk) and spacing. From this > information, you can check if the source to isocenter (point 0,0,0) > distance puts the source in or out of the volume. > I will try to correct this bug asap possible, > Simon > > > On Wed, Feb 27, 2013 at 6:45 AM, M Miller wrote: > >> Scaling the phantom worked perfectly. >> You mentioned that a reconstruction will need to specify the source >> outside of the volume. How can I specify that? What parameters would you >> pass to rtkfdk.exe to reconstruct the phantom? >> >> Thanks >> Micah >> >> --- On *Tue, 2/26/13, Simon Rit * wrote: >> >> >> From: Simon Rit >> Subject: Re: [Rtk-users] trouble projecting phantom >> To: "M Miller" >> Cc: rtk-users at openrtk.org >> Date: Tuesday, February 26, 2013, 12:36 AM >> >> >> Hi Micah, >> I have given it a try and I think I obtain the expected result if you use >> the SheppLogan file that we provide here >> http://midas3.kitware.com/midas/download?items=27326 >> However, the result might surprise you because your source is in the >> phantom (sid=65.93 vs largest ellipsoid which is A=88.32 B=115.2 C=117.76). >> You can scale the shepp logan phantom using the following command >> rtkprojectshepploganphantom -g geo.xml -o proj.mha --spacing 1 >> --dimension 1024 --phantomscale 10 >> You will then obtain more meaningful results, i.e., the full phantom will >> be in the beam. >> Note by the way that if you use rtkfdk afterwards, you will have to be >> careful to have the source outside the volume as well. I have recently >> noticed that there was a bug when the source was in the volume, I will fix >> it as soon as possible. >> Simon >> >> On Tue, Feb 26, 2013 at 2:23 AM, M Miller >> > wrote: >> >> I've been unable to project a phantom, or anything else, at high >> magnification. Here's the steps I've think should work, but don't. >> >> "rtksimulatedgeometry.exe --output=geo.xml --nproj=1600 --arc=360 >> --sdd=870.86 --sid=65.93" >> >> These are the values from my actual CT system. Next I try using >> rtkprojectgeometricphantom to create some projections. >> >> "rtkprojectgeometricphantom -g geo.xml -o projs\projections1024.mha >> --spacing 1 --dimension 1024 --phantomfile >> SheppLogan.txt" >> >> Inspecting the output projections shows they're mostly unrecognizable. >> Repeating these steps with a higher SID does produce the expected phantom >> projections. I've had no luck using real data, so I thought I'd try a >> simple phantom, but it's turned out to be not so simple. >> >> Is RTK limited somehow with respect to magnification, or are my >> expectations unrealsitic? Am I missing a command-line switch? >> >> Thanks >> Micah >> >> _______________________________________________ >> Rtk-users mailing list >> Rtk-users at openrtk.org >> http://public.kitware.com/cgi-bin/mailman/listinfo/rtk-users >> >> >> > -------------- next part -------------- An HTML attachment was scrubbed... URL: From hsieandy at gmail.com Mon Mar 11 02:32:15 2013 From: hsieandy at gmail.com (Andy Shieh) Date: Mon, 11 Mar 2013 17:32:15 +1100 Subject: [Rtk-users] About SART Message-ID: Hi Simon, I have a question regarding practical implementation of SART. As far as I know, one needs to find the ray-sum coefficient w_ij by bilinear interpolations to simulate the forward projection process. Then, the backprojection process is done in an algebraic manner using the w_ij we found according to the SART formula. However, according to a few open source codes I've found, as well as rtksart, I realize in many cases people simply use a traditional backprojection (with or without ramp filter) for the backprojection step. Is there any reference, or justification, that this is equivalent, or approximately equivalent, to the actual SART? Thanks! Not sure if I've described the question clear enough. Please let me know if it doesn't make much sense and I'll try to elaborate a bit more. Thanks :) Cheers, Andy -------------- next part -------------- An HTML attachment was scrubbed... URL: From simon.rit at creatis.insa-lyon.fr Mon Mar 11 02:56:03 2013 From: simon.rit at creatis.insa-lyon.fr (Simon Rit) Date: Mon, 11 Mar 2013 07:56:03 +0100 Subject: [Rtk-users] About SART In-Reply-To: References: Message-ID: Hi Andy, If I understand you correctly, yes there is: http://www.ucair.med.utah.edu/FacultyZeng/zeng-papers/Unmatched2000.pdf But in rtksart, you can change the backprojector to a matched one with the option --bp Joseph. Note that rtksart has not been tested in depth so let us know if you encounter problems, Simon On Mon, Mar 11, 2013 at 7:32 AM, Andy Shieh wrote: > Hi Simon, > > I have a question regarding practical implementation of SART. As far as I > know, one needs to find the ray-sum coefficient w_ij by bilinear > interpolations to simulate the forward projection process. Then, the > backprojection process is done in an algebraic manner using the w_ij we > found according to the SART formula. However, according to a few open source > codes I've found, as well as rtksart, I realize in many cases people simply > use a traditional backprojection (with or without ramp filter) for the > backprojection step. Is there any reference, or justification, that this is > equivalent, or approximately equivalent, to the actual SART? Thanks! > > Not sure if I've described the question clear enough. Please let me know if > it doesn't make much sense and I'll try to elaborate a bit more. Thanks :) > > Cheers, > Andy > > _______________________________________________ > Rtk-users mailing list > Rtk-users at openrtk.org > http://public.kitware.com/cgi-bin/mailman/listinfo/rtk-users > From seryongkang at naver.com Mon Mar 11 04:28:52 2013 From: seryongkang at naver.com (=?UTF-8?B?6rCV7IS466Oh?=) Date: Mon, 11 Mar 2013 17:28:52 +0900 (KST) Subject: [Rtk-users] =?utf-8?q?__About_SART_with_CUDA?= Message-ID: Hello , simon. I used RTK for SART with CUDA, GPU. I have used this code with geforce GTX 560, CUDA SDK 4.1,and it was working greatly. But, recently I changed my graphic card to geforce GTX 680, and then the RTK SART code have been not working. The error message is itk::ERROR:: CUDA ERROR: invalid device function. How can I fix my problem. Please help me. Thanks. -------------- next part -------------- An HTML attachment was scrubbed... URL: From simon.rit at creatis.insa-lyon.fr Mon Mar 11 06:03:16 2013 From: simon.rit at creatis.insa-lyon.fr (Simon Rit) Date: Mon, 11 Mar 2013 11:03:16 +0100 Subject: [Rtk-users] About SART with CUDA In-Reply-To: References: Message-ID: Hi, This can happen when the requested ressources are not in accordance with the GPU characteristics, e.g., wrong block sizes, texture size, etc. Try to locate the line giving you this error (it's given in the exception error message) and modify the hard coded parameters. We can help if you provide the complete error message. Simon On Mon, Mar 11, 2013 at 9:28 AM, ??? wrote: > Hello , simon. > > > > I used RTK for SART with CUDA, GPU. > > > > I have used this code with geforce GTX 560, CUDA SDK 4.1,and it was > working greatly. > > > > But, recently I changed my graphic card to geforce GTX 680, and then the > RTK SART code have been not working. > > > > The error message is itk::ERROR:: CUDA ERROR: invalid device function. > > > > How can I fix my problem. Please help me. > > > > Thanks. > > _______________________________________________ > Rtk-users mailing list > Rtk-users at openrtk.org > http://public.kitware.com/cgi-bin/mailman/listinfo/rtk-users > > -------------- next part -------------- An HTML attachment was scrubbed... URL: From simon.rit at creatis.insa-lyon.fr Mon Mar 11 08:13:31 2013 From: simon.rit at creatis.insa-lyon.fr (Simon Rit) Date: Mon, 11 Mar 2013 13:13:31 +0100 Subject: [Rtk-users] About SART with CUDA In-Reply-To: References: <911ac3fb64426d877dccaebe325fd6d2@tweb03.nm.nhnsystem.com> Message-ID: Please continue using the mailing list. It's hard too say, maybe the volume is too big or too small for your GPU. Have you tried doing "make test" to see if the problem is related to this particular reconstruction or if this is a more general problem? What is the size of the projection images and the volume that you use? Simon On Mon, Mar 11, 2013 at 11:52 AM, ??? wrote: > Hi. > > > > This is the line giving this error. > > > > : In rtkCudaForwardProjectionImageFilter.cu line : 251 >> error > message is invalid texture reference > > > > // Set texture parameters > > tex_vol.normalized = false; // access with > normalized texture coordinates > > tex_vol.filterMode = cudaFilterModeLinear; // linear interpolation > > tex_vol.addressMode[0] = cudaAddressModeClamp; // clamp texture > coordinates > > tex_vol.addressMode[1] = cudaAddressModeClamp; > > // Bind 3D array to 3D texture > > cudaBindTextureToArray(tex_vol, (cudaArray*)dev_vol, channelDesc); > > CUDA_CHECK_ERROR; --------------->>> this is line 251 > > > > > > and : > > > > In rtkCudaBackProjectionImageFilter,.cu line : 251 >> error message is > invalid device function > > > > kernel <<< dimGrid, dimBlock >>> ( dev_vol, > > make_int3(vol_dim[0], > vol_dim[1], vol_dim[2]), > > blocksInY, > 1.0f/(float)blocksInY ); > > } > > CUDA_CHECK_ERROR; ------------>>> this is line 251 > > > > > > I can't understand about this error..... > > > > Please help me, simon. > > > > > > Thanks. > > -seryong > -------------- next part -------------- An HTML attachment was scrubbed... URL: From hsieandy at gmail.com Mon Mar 11 19:29:05 2013 From: hsieandy at gmail.com (Andy Shieh) Date: Tue, 12 Mar 2013 10:29:05 +1100 Subject: [Rtk-users] About SART In-Reply-To: References: