[Rtk-users] Trouble Reconstructing 3D CBCT using 4D DVF
Matthew J. Riblett
riblettmj at mymail.vcu.edu
Tue Feb 3 10:36:03 EST 2015
Simon,
I think that it makes it past SetFrame and crashes on Update. Just to be sure, I put a debug message after SetFrame and it does make it to that point, so its got to be on the Update line.
— Matt
__
Matthew J. Riblett
Virginia Commonwealth University
Department of Radiation Oncology
Medical Physics Graduate Program
Office: Sanger Hall, Room B1-013
401 College Street | P.O. Box 980058
Richmond, Virginia 23298
VCU Email: riblettmj at vcu.edu <mailto:riblettmj at vcu.edu>
MCV Office Phone: +1.804.628.4858
> On Feb 3, 2015, at 2:22 AM, Simon Rit <simon.rit at creatis.insa-lyon.fr> wrote:
>
> That's weird because SetFrame does nothing else than change the value
> of the member Frame. Are you sure it's on SetFrame that the crash
> occurs?
>
> On Mon, Feb 2, 2015 at 7:44 PM, Matthew J. Riblett
> <riblettmj at mymail.vcu.edu> wrote:
>> Simon,
>>
>> I rebuilt the rtkfdk app (actually all of RTK) in ‘Debug’ mode and with the
>> added debug line you suggested and it doesn’t look like the pointer is null.
>> Running it through gdb, I got the following info:
>>
>> m_Deformation=0x16ff820
>>
>> (gdb) x 0x16ff820
>> 0x16ff820: 0x011c51d0
>> (gdb) x 0x011c51d0
>> 0x11c51d0
>> <_ZTVN3rtk28CyclicDeformationImageFilterIN3itk5ImageINS1_6VectorIfLj3EEELj3EEEEE+16>:
>> 0x005fb874
>> (gdb) x 0x005fb874
>> 0x5fb874 <rtk::CyclicDeformationImageFilter<itk::Image<itk::Vector<float,
>> 3u>, 3u> >::CreateAnother() const>: 0xe5894855
>> (gdb) x 0xe5894855
>> 0xe5894855: Cannot access memory at address 0xe5894855
>>
>> I’m still waiting on the transformix DVF, but I’ll give that a try too when
>> its done.
>>
>> Thanks again!
>>
>> — Matt
>>
>> __
>> Matthew J. Riblett
>> Virginia Commonwealth University
>> Department of Radiation Oncology
>> Medical Physics Graduate Program
>>
>> Office: Sanger Hall, Room B1-013
>> 401 College Street | P.O. Box 980058
>> Richmond, Virginia 23298
>>
>> VCU Email: riblettmj at vcu.edu
>> MCV Office Phone: +1.804.628.4858
>>
>>
>>
>> On Feb 2, 2015, at 1:16 PM, Matthew J. Riblett <riblettmj at mymail.vcu.edu>
>> wrote:
>>
>> Thanks for the quick reply Simon! I am using the rtkfdk app, and I will go
>> ahead and check to see if the pointer is null and let you know. My current
>> check is having transformix to generate a smaller DVF to see if that helps,
>> so the server is getting that done as we speak.
>>
>> You might also like to know that I’ve got RTK running on a Cray CX-1 (baby
>> supercomputer) sitting next to the aforementioned Sun x4600-M2. :)
>>
>> — Matt
>>
>> __
>> Matthew J. Riblett
>> Virginia Commonwealth University
>> Department of Radiation Oncology
>> Medical Physics Graduate Program
>>
>> Office: Sanger Hall, Room B1-013
>> 401 College Street | P.O. Box 980058
>> Richmond, Virginia 23298
>>
>> VCU Email: riblettmj at vcu.edu
>> MCV Office Phone: +1.804.628.4858
>>
>>
>>
>> On Feb 2, 2015, at 12:19 PM, Simon Rit <simon.rit at creatis.insa-lyon.fr>
>> wrote:
>>
>> Hi,
>> You're the first RTK user I know to use it on Sun! Nice!
>> If you've got 64 GB ram, that should be enough... although you could
>> ask elastix to generate a smaller DVF, RTK linearly interpolates
>> between vectors. You could even use B-spline coeffs directly but you'd
>> have to create a new DVF.
>> My guess is that there is a problem with the initialization of
>> m_Deformation since the line you mention is the first time it's ever
>> used in the class. Do you use the rtkfdk app? I would start by
>> checking that the pointer is not null before this line, sg like:
>> std::cout << "m_Deformation=" << m_Deformation.GetPointer() << std::endl;
>> If it equals 0, then there is something going with the deformation.
>> You can check it in the main app, rtkfdk.cxx.
>> Simon
>>
>> On Mon, Feb 2, 2015 at 6:03 PM, Matthew J. Riblett
>> <riblettmj at mymail.vcu.edu> wrote:
>>
>> Hello RTK Users,
>>
>> I’m just getting started using RTK as part of my graduate research and was
>> hoping that someone might be able to help me sort out an issue I’ve been
>> having reconstructing a 3D CBCT image dataset while applying a 4D
>> deformation field — I’m reconstructing with ~2400 projections and the 4D DVF
>> has been generated by transformix (.mhd format). I followed, for the most
>> part, the MCCBCTReconstruction example on the RTK Wiki.
>>
>> While reconstructing, rtkfdk reads in all of the projections to memory,
>> begins the ‘Reconstructing and Writing’ phase and proceeds for a few seconds
>> before hanging — it doesn’t exit or fault, just drops to 0% CPU while
>> retaining the projections in memory.
>>
>> By writing some basic debug messages into the source code, I’ve been able to
>> determine that the issue lies somewhere at m_Deformation->Update() [Line 99
>> of code/rtkFDKWarpBackProjectionImageFilter.txx]. It appears that thread 0
>> makes it past the m_Deformation->SetFrame(iProj) line, but never makes it
>> past the subsequent update. I’m wondering if this has anything to do with
>> the size of the DVF generated by transformix: approximately 4.0GB
>> (512x100x512x10 image, 4-channels, FLOAT).
>>
>> Just to provide a little more background information, I’m running RTK on a
>> 32 core Sun Server (8x Quad-Core AMD Opteron 8356, 2.3Ghz) with 64GB of
>> memory running Ubuntu 14.04. RTK was built using the default ‘Release’ make
>> configuration and linked to ITK 4.5.2. Elastix and Transformix are both the
>> latest version, 4.7.
>>
>> Has anyone else run into a similar issue with reconstructing a similar
>> image, or does anyone have any thoughts as to how to resolve this sort of
>> issue?
>>
>> Thanks in advance!
>>
>> — Matt
>>
>> __
>> Matthew J. Riblett
>> Virginia Commonwealth University
>> Department of Radiation Oncology
>> Medical Physics Graduate Program
>>
>> Office: Sanger Hall, Room B1-013
>> 401 College Street | P.O. Box 980058
>> Richmond, Virginia 23298
>>
>> VCU Email: riblettmj at vcu.edu
>> MCV Office Phone: +1.804.628.4858
>>
>>
>>
>> _______________________________________________
>> 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/20150203/4d748108/attachment-0009.html>
More information about the Rtk-users
mailing list