BDSIM for proton therapy gantry simulation Accelerator Group

BDSIM for proton therapy gantry simulation Accelerator Group Christmas Meeting Andrew Green

About Me • Recently finished my Ph. D (2012 – 2016) • Did my undergrad @ Manchester too (2007 – 2012) • Lots of work on proton therapy – Specifically Monte Carlo • Now work @ Christie in Marcel van Herk’s group

Proton Therapy • Uses protons to treat tumours • Protons stop – Most of the time in the right place • Should be able to spare quite a lot of normal tissue. • However… Obligatory Bragg Peak plot. All talks discussing proton therapy are required by law to have at least one plot showing how much better than X-rays protons are.

Proton therapy • Protons don’t always go where you expect them to. • Monte Carlo will be needed for treatment planning. • But… • . . . it’s far too slow. Pencil beam algorithm (left) vs Monte Carlo (right). The arrow indicates a range difference due to multiple Coulomb Scattering. Adapted from Paganetti (2008)

My Work • Tried a few tricks to make Monte Carlo faster – Supercomputer – Intel Xeon Phi – Modified transport algorithm – Cloud computing • Also developed a different technique for dose comparisons.

The Woodcock Algorithm • Took the full 4 years to develop • Seemed like a good idea at the time • Discovered two things: – Most tracking time is for secondaries – Depositing CSDA losses properly is really hard

Thesis • Hardest part of Ph. D • One piece of advice: – Start writing early. – Even if it’s shit. • It can be quite fun when you get into it

Thesis

The title said something about BDSM. . • BDSIM is a wrapper for Geant 4 to do accelerator tracking • We’ve been tracking through a research beamline which is going to be installed at the Christie proton therapy centre • Also looking at tracking through an RF cavity to look at losses.

Tracking losses through the cavity (in progress!) • Previously used a ‘homebrew’ tracking model plus iris radii – Gives approximate beam loss, but not subsequent radiation shower • Working on a combined beam tracking/loss calculation in GEANT 4 – Uses BDSIM framework – We added CAD import – Field map import in progress – We added improved tracking algorithm – We added some parallelisation tools – A good general tool for proton therapy centre design, e. g. losses, shielding etc.

Optics Validation • Does BDSIM correctly model the focusing of the beam? – Use well known accelerator code MADX to get optics functions. – Try to replicate them in BDSIM • Looks good!

Optics Validation • Not everything works out of the box – Something odd going on with units? • Pretty sure the beam isn’t 1. 5 m! • Should be fixed fairly easily.

Loss Maps • CAD import linac structure: – For now, simulate separate to other geometry The cavity was 10 x too small! • BDSIM uses ROOT to collect data • Easy(ish) to produce histograms of primary loss • Very very preliminary!

The Future • Tracking through a gantry – We will need a gantry optics file • Not been done before • Can we track from the accelerator into the patient? – Include degraders, beam transport and gantry, then deposit dose in the patient • Probably do a mashup of everything used so far