Minutes

Friday, July 30 2010

Other versions of Varian OBI have a different file format (header)
Long term plan: improve geometric calibration. Greg has a phantom and associated software. It gives you 6 DOF/projection.
Greg wil have Elekta cone-beam by fall 2011.
Geometry XML file will contain and no matrix: source position, detector position, detector size, tomography position, tomography size

- Synergy reader: test the benefit of using a log lookup table (Simon)
- Develop the CPU version of the backprojection (Simon)
- Proposal of geometry file for the next meeting (Simon)
- Look into geometric calibration of projections images in Varian (Greg)

TODO
- Varian Obi: where is the log taken in Plastimatch? (Greg)
- Try to find out what the constructors are doing before storing the raw data. (Greg, Simon)
- Send the scatter correction algorithm and ghosting correction to Greg (Simon)
- Run a scan of the CatPhan at the same dose (Simon and Greg)
- Look into OpenCL FFTW (Greg)
- Looking at other solutions (Greg, Simon)
- Generic framework for reconstruction like in ITK for registration (David)

Thursday, September 9 2010
TODO
- Geometry ok. Add orientation of projections with respect to volume?
- Strip out xbase for what’s required (Simon)
- Link with Plastimatch (Grep)
- Cupping: truncation + scatter…
- Raytracer. Greg will try to convert the plastimatch one to itk style.
- FDK. Simon will finish the algorithm and optimize the backprojection.
- Greg will send me the Varian maintenance manual for geometric calibration. (DONE: 2010-09-09)

Friday, November 5 2010
Participants: Greg, Naga, Tony, David, Simon
- a way to share code between Plastimatch and RTK must be set as soon as possible. A solution would be to use an “external repository” which would embed RTK in Plastimatch. It would act like a symbolic link and Plastimatch users would automatically download both repositories.
→ Simon will investigate the bridge between svn and git, i.e. download a git rep using svn
→ Greg will test the external rep solution
→ The issue of common files (e.g. cmake files) need to be solved. Two potential solutions: move the files from Plastimatch to RTK or use a third repository. The first solution sounds better…
- licensing: current file https://github.com/SimonRit/RTK/blob/master/LICENSE.txt looks ok but it has been suggested to add an explicit non clinical purpose and to replace copyrights with a consortium.
When required, third party code will be added to the repository and the source will be added to a copyright file, like in Plastimatch.
- GPU development: Plastimatch has OpenMP, Cuda and OpenCL versions, the Cuda version being the most mature. v1.0 of FDK in RTK could be a wrapping of Plastimatch code in an ITK filter. Simon will try to start doing it.

Monday, November 29 2010
Participants: Greg, Naga, Tony, Simon
- Create an RTK account for Tony. Simon will contact Tony about it. (DONE: 2010-11-30)
- License / Copyright: each author retains copyright. Suggestion: file in the distribution acknowledging third party projects, like COPYRIGHT.TXT in Plastimatch. Simon will create it. (DONE: 2010-11-30)
- GitHub: bug when moving file? Need to be tested.
- CUDA / OpenCL utils: keep copies in RTK, Plastimatch is the master. (files removed, functions in itkCudaUtilities.cu DONE: 2010-11-30)
- Time plastimatch to compare with rtkfdk (Simon)
- Try cufftw (Simon)
- Add degrees of freedom to itkGeometry.h (Simon)
- Try 2d cuda texture (Simon) (better/faster, DONE: 2010-11-30)
- Try plastimatch opencl code (Simon)

Monday, July 27 2011
Participants: Greg, Simon
- Greg is working with Salzburg group (P. Steininger, M. Neuner and H. Deutschmann) on 2D/3D registration. There is a copy of the code in the Plastimatch repository, see also http://www.insight-journal.org/browse/publication/776). The code currently requires a NVidia graphic card, CPU implementation is a priority.
- Geometry file review: more information on context and schemes are required for understanding.