Characterisation of the electron spectrum of a therapeutic linear accelerator for dosimetry and treatment planning next up previous

Characterisation of the electron spectrum of a therapeutic linear accelerator for dosimetry and treatment planning

Richard Hugtenburg1,2 and Zaizhe Yin2

1Imaging and Medical Physics Group
Queen Elizabeth Medical Centre
University Hospital Birmingham NHS Trust
Birmingham
B15 2TH
United Kingdom

2School of Physics and Astronomy
University of Birmingham
Birmingham
West Midlands, B15 2TT
United Kingdom

Electronic correspondence:
r.p.hugtenburg@bham.ac.uk
http://web.bham.ac.uk/r.p.hugtenburg

























Abstract:

Introduction: High dosimetric accuracy in radiation oncology is essential to maximise the probability of tumour control (TCP) and minimise the probability of normal tissue complication (NTCP). It is usual to combine or compare measurements to obtain higher overall precision. Combining data from independent measurements requires precise modelling of the response of the detector systems and a means to correlate these results with dose in vivo.

Methods: The Monte Carlo (MC) method enables information describing the source, detector and their environments to be used to compute, with high accuracy, quantities of dosimetric significance. MC has the potential to be an extremely useful tool in radiation dosimetry, but this potential is often unrealised because fundamental information about the properties of the source and detectors, upon which MC calculations are strongly dependent, are not easily obtainable. As such, the conventional MC method cannot fulfil the requirements of this modelling system, as it is unable to provide correlated expectations of quantities without accurate information about the initial configuration of the system. In practice, workers employ an iterative approach which, if performed manually, can be tedious, but is also potentially inaccurate if the space of degrees of freedom in the model is complex. A Bayesian approach, in particular, the Markov chain Monte Carlo (McMC) method, has been used to automate sampling from uncertainty distributions in the prior domain and enable measured, posterior, data to constrain the source model. Results: The MCMC method has been applied to a simple source model required for the dosimetry of total skin electron radiotherapy and in vitro radio-carcinogenesis studies, as well as more complex source models required for conventional radiation treatment planning (RTP). This work demonstrates that detailed descriptions of the source are unnecessary to achieve accurate calculations of dosimetric quantities, such as mass-collision stopping powers and the geometry and field size dependence of dose and that generalised models of a therapeutic linear accelerator in combination with normal commissioning data can be used to achieve a 3%/3mm criterion for dosimetric accuracy.

Conclusion: Accurate radiation dosimetry demands that all information relating to the modelled system be used, including a quantification of its limitations. MCMC bridges the gap between detailed MC simulations and traditionally