1 / 18

Sources and Modelling for Proton Computed Tomography

Sources and Modelling for Proton Computed Tomography. School of Physics and Astronomy, University of Manchester & Cockcroft Institute for Accelerator Science and Technology. Hywel Owen, Andrew Green, David Holder. UK Proton Therapy Centres - Update. Timeline

gabe
Download Presentation

Sources and Modelling for Proton Computed Tomography

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Sources and Modelling forProton Computed Tomography School of Physics and Astronomy,University of Manchester & Cockcroft Institute for Accelerator Science and Technology Hywel Owen, Andrew Green, David Holder

  2. UK Proton Therapy Centres - Update • Timeline • Pre-qualification questionnaire has been issued • Technical specification has been issued:technical advice from PSI (Lomax), UPenn (Maughan), UCL, CI;very useful exercise in learning detailed specs to design against • Competitive dialogue during 2014 • Manufacturer chosen/ contract placed end 2014this is when we know what kind of machine it will be • First patients 2018 • Specification: • 2 centres • 230/330 MeV protons only, spot scanning • 2 Gy/min/litre • 1500 patients (750/centre), 3 or 4 treatment rooms per centre

  3. Christie Site

  4. UCLH Site Cruciform Building Rosenheim Building Odeon Site Jeremy Bentham Pub Macmillan Cancer Centre Spearmint Rhino Simon Jolly, University College London

  5. Cockcroft/Christie research activities on protons • Manchester/Christie: 3 postdocs+ students (2 PhD, 2 MSc) • Aitkenhead/Richardson/Charlwood: Treatment planning, spot models, throughput, dosimetry, delivery methods (e.g. penumbra control) • Holder: Gantry design • Garland: Design of medical FFAG • Green: Monte-Carlo tracking and verification • Liverpool/Clatterbridge • Silicon beam halo monitor • Silicon tracker/calorimeter for proton tomography • Medical FFAG design

  6. UK Research Activities and Interests in Radiotherapy • Manchester – FFAG design, gantry design, Monte Carlo, computational methods, Si dosimetry, novel detectors. • Liverpool – diagnostic instrumentation, neutron instrumentation • Lancaster – high-gradient RF cavities • National Physical Laboratory – dosimetry standards for protons • Clatterbridge – new nozzle design, backgrounds • UCL/UCLH – dosimetry, neutron instrumentation, novel imaging • Oxford – radiobiology, throughput/computation • RHUL – Beam tracking/background estimation • Imperial – FFAG gantries, laser-plasma protons and manipulation, radiobiology • Huddersfield – FFAG design, beam tracking/space charge • STFC RAL – laser proton acceleration, laser isotope production, FFAG design • STFC DL – conventional magnet design

  7. How many protons? • Use example PTV 10x10x10 cm • 1 litre • Shallowest depth 10cm; 112 MeV • Deepest depth 20cm; 166 MeV • Most of the 166 MeV proton dose goes into PTV • Little of the 112 MeV proton dose goes into PTV • About half of dose to PTV • ~22 pJ per proton • 45 Gp for 1 Joule • 90 Gp/16 nC for 1 Gy in 1 litre PTV • 1 min -> 750 Mp/s = 0.12 nA • GEANT4 simulation of IMPT • 90.8 Gp/Gy in PTV • 43.1% in PTV • 43.7% upstream (proximal) • 0.58% downstream (distal) • 12.5% lateral

  8. Manchester/Christie Work • Concentrating on improving proton tomography • Technical rather than clinical focus • Improved source: • 330 MeV proton FFAG • Normal-conducting and superconducting designs under investigation • Improved gantry • Scaled NIRS gantry design for compact treatment room • 330 MeV design in same space as normal gantry • Dose verification and diagnostic instrumentation • Improving Monte Carlo calculations of treatment plans

  9. ‘Prototypical’ Isocentric Gantry – PSI Gantry 2 250 MeV protons ~ 38 cm in water, 2.46 Tm rigidity, NC B<1.6 to 1.8T

  10. Magnetic Rigidity • E.g. Heidelberg 425 MeV/u • Br = 6.57 Tm • 1.8 T -> r = 3.65 m • 3.3 T -> r = 2 m Cryocooler Connections (no fluid) CEA/IBA/Etoile

  11. NIRS Gantry • NIRS design is for 430 MeV/u carbon • Our design is for 330 MeV proton • Proton optics design complete • Now tracking through apertures • Magnet design next NIRS (Japan) 3.0 T for 430 MeV/u 200 t total 13 m x 5.5 m

  12. Monte Carlo Validation • MC Validation is a niche area compared to other applications • Many voxels, regularly spaced • General codes exist: • MCNP/X, Fluka, GEANT4 • Lots of optimisation has already been done in general codes to increase speed • Voxel navigation • Efficient memory usage • Efficient algorithms for particle transport • In-field dose computation possible with GPUs, e.g. gPMC • Drop physics: • Continuous slowing down model; empirical straggling • No electron transport • Simplified secondary transport, e.g. no neutrons • Out-of-field dose computation needs to retain the physics • Out-of-field neutron generation and dose are important, e.g. in pediatric secondary tumour induction

  13. Validating a Dose Calculation • Monte Carlo with sufficient physics, e.g. to model effects of implants, bones etc. • Validated range calculation • Requirement for benchmarking • NPL benchmarks to be developed • Must agree with experiment! • Sufficient number of protons for out-of-field dose error to be quantified • Probably about 10^7, but needs study • Must be fast • About 1 hour to begin with • Use same code for range verification for proton CT • Quantitative comparison

  14. What sort of computer? • GPU • Not well suited to particle transport (track lengths differ) • Supercomputer (e.g. Hartree Centre, 112000 CPUs) • Great for fast ‘one-off’ calculations • Moving data to/from system • Patient data security • GRID-based • Moving data • Security • Small clusters, e.g. 48-core, 96… • MIC (Many-Integrated Core) – Xeon Phi • Many CPUs on a card • Physically small, low power – fits in the planning suite! • Just released (Nov 2013) • Examine prototype system to see comparative speed

  15. Xeon Phi • 61 cores = 244 Threads maximum • 1.1GHz clock speed • 16GB RAM – currently a limitation • Optimized for highly vectorised code • Theoretically – 1TFLOP (ish)

  16. Xeon Phi vs AMD Opteron

  17. Xeon Phi and GEANT4 • GEANT4MT just released • Multithreaded, can utilise shared memory in Xeon Phi • Compiled and running happily on test system • Considerable benefits over GPGPU accelerators: • Simple compilation: just add the –mmic flag • No need to convert to proprietary languages (eg CUDA) • About the same price as a top-end GPGPU • More memory (?) • Test Xeon Phi system (2 cards) bought by UOM • 1 card in use so far • Test system at Christie Hospital about to be turned on • Bids made for 10 more Xeon Phi cards

  18. Geant4 Dose output • Developed code which can output dose grid from Geant4 • Needs work… • Very quick: • 10^7 protons • 1.6 hours (on Opteron) • ~1.2 hours (on Xeon Phi per card) (t.b.c.) • Could be used for validation?

More Related