Essay: Medical Radiotherapy

This project is part of the framework of medical radiotherapy. It focus on the method of evaluating of
the Dose Convolution Integrals for a Pencil Beam Model for proton therapy. The method we are going
to use is the Fast Gaussian Transform (FGT).
The current planning system used for proton therapy uses a simple ray- casting model of the proton
beam. While this model is adequate for current treatments at iThemba LABS, it cannot accurately
model a number of prop- erties of the beam. In particular, this model does not accurately capture the
penumbra of the beam, nor can it model devices in the beam line very accurately. Consequently iThemba
Labs is limited to using treatments that involve comparatively few treatment beams, and cannot use
more complex beamline elements, such as compensators, to shape the beam.
This inadequancies can be addressed by using a pencil beam model, which more accurately models
the physics of the beam line. A pencil beam model requires considerable work on accurately mod-
elling the iThemba LABS proton beamline, and effecient evaluation of the model poses a num- ber of
computational changes.
In particular, the pencil beam model not only requires very fast ray-tracing routines, it will also require
extremely efficient algorithm for the 2D dose convolution integral, since a straightforward application of
the Fourier or Hartley Transforms will not be suitable for this application. This is even more of an issue
for the extended pencil beam models, which are more accurate, but considerably more omputationally
very expensive since ray- tracing calculations along a bundle of lines are required for each pencil beam,
and more evaluations of the dose convolution intergrals are required.
The aim of this project is to investigate methods for efficiently evaluating the dose convolution inte-
grals, with methods suitable for parallel implemen- tation on a GPU of particular interest and propose
algorithms which can be used to help develop a pencil beam model for the iThemba LABS proton
beamline.