CCPi

Staffing: The project is proceeding as planned has a strong links with core developers Barry Searle (based at DL), Srikanth Nagella and Ronald Fowler (based at RAL). Over the last year due to an underspend we have had 0.2 FTE Martin Turner and 0.45 FTE Erica Yang.

Administration: Working group meetings were organised during Q2 and Q4 2014: A series of developers workshop have continues on a quarterly basis: 14th Jan 2014and 20th May 2014 in Birmingham; 22 July 2014 at Nottingham and 22 October 2014 in QMUL. We are also assisting in the second Tomography for Scientific Advancement symposium (ToScA) meeting in the Natural History Museum (1-3 Sep 2014). A monthly showcase involving software show-and-tells were organised at RAL has continued and there are 18 events in the series; including at the end two recorded sessions (47 external views).

Collaboration: contact has continued between CCPi and CCP-ASEArch. A CCPi Seminar was held Nov 2013 and starting in April 2014 was work by Evgueni Ovtchinnikov; involving numerical methods, to apply a multigrid solution to improve the issues caused by gradient based regularization coefficients in order to accelerate and stabilise convergence. This has now resulted in a publication to be submitted after collaboration with Daniil Kazantsev (CCPi Flagship Fellow)..

Publications: Related to the code integrated, with the flagship project, within the CCPForge, three papers have been accepted by Daniil Kazantsev including attendance at the Applied Inverse Problems conference, 2015 in Helsinki.; and the Royal Society Phil Trans A special edition now has a complete set of titles and authors (nine), with seven under peer review: Theme Issue is titled "X-ray Tomographic Reconstruction for Materials Science".

Development targets met:

  • Initial particle tracking code based on MXIF is complete and in use. Modifications will be made to generalise function to determine proximity of particles from one form to the next.
  • We are implementing histogram based thresholding methods including those from ITK, and further tools are ongoing.
  • Iterative reconstruction development proceeds; we have completed a NeXus? reader for Diamond Light Source data sets; and have run test Diamond data sets on the Hartree Centre ScaleMP? node (order 1TB RAM), where iterative methods require about 20hrs/iteration. The code has been modified to reduce memory requirements for workstations and for the following stage required optimisation to reduce compute cost per iteration which is currently ongoing.
  • The external profile of the project has been raised by updating the CCPi website and creating a dedicated CCPi YouTube? channel:http://www.youtube.com/channel/UCGB578xcyXNQyiBsufEFIeQ
  • Modify CGLS code to analyse example Diamond data set complete by Q3 2013.
  • Optimise projection algorithm at the core of iterative routines for an accelerated architecture (GPU/Phi) Q1 2014. The optimised Diamond version has been tested on Xeon Phi and further work tests will occur as part of the ongoing optimisation.
  • Added two new iterative reconstruction algorithms from network members or flagship project to reconstruction suite (Q1 2014). At a developers' meeting in Manchester in March it was discussed how the developer/flagship PDRA should integrate their algorithms into the CCPForge archive and a way forward is being used.

Metrics 2013/4 need values for 2014/5

FOR ALL CCPs. We would highlight that during the financial year 2013/4 (figures from last year’s report in brackets):

  • STFC staff have produced some 51 (107) publications not including papers in preparation
  • STFC staff have given some 101 (109) presentations at meetings ranging from informal site visits to invited talks at major conferences.
  • There are some 1629 (1497) publications that reference codes developed through the SLA
  • Some 4451 (3732) training days covering !PhD supervision, code workshops and scientific meetings.

NEED metrics for CCPi.

April 2014- March 2015



Technical Appendix plans for 2015

CCPi – Tomographic Imaging

Non destructive 3D X-ray imaging is becoming increasingly important in many areas of science and its application to Energy, Healthcare and Security. For example it is having a dramatic impact on fields as diverse as security (e.g. baggage and body scanning at airports and screening of vehicles at ports), engineering (e.g. visualising stress corrosion cracking in nuclear plant and the degradation of fuel cells) and medicine (e.g. cancer treatment and artificial tissue engineering). The spatial and temporal resolutions are increasing dramatically. RC funded synchrotron sources are rapidly increasing the numbers of x-ray imaging instruments available (the European Synchrotron Radiation Facility (ESRF) now has 10 beamlines, Diamond Light Source (DLS) is currently building 4 new imaging beamlines). Laboratory x-ray imaging facilities are becoming increasingly widespread. This expansion is mirrored elsewhere with the global CT market now worth $150M (+ $5B in medical CT) both expanding at 10% pa , while 30% of the data stored on the world's computers are now medical images. Unsurprisingly, papers on x-ray tomography have also increased sharply this decade (doubling every 3 years with 1,500 published last year).

This CCP aims to provide a toolbox of algorithms that increase the quality and level of information that can be extracted from computer tomography. The CCP will bring together the imaging community, maximise return on investment in software development and ensure longevity, sustainability and re-use of code.

Outline of Proposed Core Support Activities:

Most people who acquire and analyse X-ray images have little or no knowledge of X-ray tomography reconstruction and analysis, consequently the involvement of the Core Support is critical if we are to achieve a significant step jump in the level of information obtained by UK science from X-ray imaging. As well as administrative support for the CCP (organization of working group meetings, maintenance of web sites), the core support team (1.5 FTE per annum) at DL will:

  • Standardise input and output data formats;
  • Standardise software coding rules and testing;
  • Assist with porting, parallelisation and optimisation on different hardware platforms including novel architecture systems.
  • Provide centralised distribution and controlled releases for software;
  • Provide installation user support for software, and run workshops/training events;
  • Collate and distribute existing algorithms and code, including licensing requests;
  • Encourage CCP participants to make their algorithms and code available;
  • Create a gallery of test real x-ray image test object datasets for mathematicians to test and benchmark new algorithms; and
  • Maintain documentation.

This project aims to provide toolboxes for both 3D image reconstruction and 3D image analysis algorithms. In simple terms it is estimated that the project requires 0.5FTE for maintaining network, website, running workshops and training course, benchmarking, licensing issues etc; 0.5FTE for developing and maintaining the 3D image analysis toolbox, 0.5FTE for developing and maintaining the image reconstruction toolbox.

CCPi Milestone Date

  • 1 Organise working group meeting/ ToScA involvement, and monthly show-and-tell sessions - Ongoing
  • 2. Support current training courses and organise developer workshop to teach (for C++ includes; Basics of an application (Avizo, ImageJ); Programming (C++/Java) with reference to particular API; API for Avizo, VolView, ImageJ; Development with ITK; Implement a module to take away) - Q4 2014
  • Port 3 or 4 iterative codes so that they can be compared on the same datasets
  • Extra targets: 3.1 Tidy up of centre of rotation issues and optimisation of iterative code with Diamond NeXuS data sets for integration with DAWN. Q4 2014 and 3.2; Port of Manchester filtered back projection into NeXuS? package as an option with DAWN - Q1 2015) Q4 2014
  • Iterative code up and working for the Nikon XTek X-Ray CT stage. Links to Nicola Wadeson and Daniil Kazantsev work - can tackle problems with fewer images.
  • Work with Tristan Lowe in Manchester to define a beta version (in the ccpforge) - (Release data release version Q1 2015) Q2 2014
  • 5. Accessible volume code (originally Matlab from Sheng Yue) has been re-implemented in !VolView and Avizo. There are scalability issues for up to 2000^3 to 4000^3 - Q3 2014
  • Implementation of two/three further quantification algorithms for and from the community (Extra item 7.1 Potential to support Avizo, !ImageJ and !ParaView. Note simpleITK package is supposed to allow ITK functionality to be called from Java so it might be possible to have common packaging of the algorithms) - Q3 2014
  • Capture an idealised dataset - suggested minimum of 2048 images each comprising 8 frames so that we could for example compare reconstructing 2048 1 frame projections with 1024 (each 2 frames summed) projections with 512 (each 4 frames summed) projections with 256 (each 8 frames summed) projections. Possibly with Diamond (Mark Bashham) - Q4 2014
  • Iterative code for the Nikon XTek X-Ray CT Released Q1 2015

Task Effort

  • Secretariat Coordination 0.2
  • Basic maintaining network, website, organising administration for workshops and training courses, 0.175
  • benchmarking, licencing issues 0.5
  • CCPi Framework Developing and maintaining the 3D image analysis toolbox 0.5
  • CCP_petem Developing and maintaining the image reconstruction toolbox 0.5
  • Evaluation of neutron imaging software 0.25 (Yang)

Total 2.5 FTEs

Staffing Effort

  • Martin Turner 0.2
  • Barry Searle 0.5
  • Srikanth Nagella 0.5
  • Ron Fowler 0.5
  • Ron Hate 0.625
  • Returned to Core SLA activities website etc. 0.175

Total 2.5

-- MartinTurner - 09 Jan 2015

Edit | Attach | Watch | Print version | History: r14 | r4 < r3 < r2 < r1 | Backlinks | Raw View | Raw edit | More topic actions...
Topic revision: r1 - 09 Jan 2015 - MartinTurner
 
  • Edit
  • Attach
This site is powered by the TWiki collaboration platform Powered by Perl This site is powered by the TWiki collaboration platformCopyright & by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback