Development of an Independent Monte Carlo Dose Calculation Tool for Validation of the Monaco Electron Treatment Planning System

Abstract

Monte Carlo (MC) is rapidly becoming the preferred algorithm for radiotherapy treatment planning systems (TPS) to obtain the most accurate dose predictions. Commercial MC TPSs rely on a number of simplifications to allow doses to be calculated in a clinically relevant timeframe. These simplifications are often out of the user’s control and may have implications on dose calculation accuracy in certain scenarios. The purpose of this work was to develop an in-house electron MC toolkit that would allow for independent validation of a commercial electron MC TPS and an understanding of the limitations of the commercial TPS. The accuracy of any MC model depends on the ability to accurately model what is present in real life. One crucial and unknown component of the model that was given priority in this work is the electron spectrum striking the exit window of the linear accelerator. A mono-energetic incident particle beam is not a true representation of the real scenario and cannot achieve sufficient accuracy for MC to be considered a reference for validating a commercial TPS. An optimized incident electron spectrum striking the exit window of the accelerator was determined by weighting mono-energetic electron energies in the form of a continuous distribution. The optimum spectrum for a given therapy beam energy was determined by minimising the difference between simulated and measured percent depth dose (PDD) data from Elekta Synergy and Agility therapy accelerators for multiple fields and source to surface distances (SSD). Spectra were initially determined without applicators present and thereby removed a significant variable in modelling. The accelerator head components that are shared between beam energies were kept constant to ensure accurate representation. Results using optimized energy spectra matched measured PDDs to within 1%/1 mm except for the first 5 mm. Measured and calculated profiles and output factors were within 2% for all fields and five energies. Spectra were then applied with electron applicators present with resulting PDDs maintaining the same accuracy. Profiles were within 1%/1 mm agreement in the clinical field for SSDs 100 and 110 cm and applicator factors to less than 3%. MC simulation profile results showed improvement over Elekta Monaco 5 TPS electron models particularly in large fields and the build-up region. Improvements were also observed when simulating dose on CT datasets due to the user’s greater ability to control voxel sizes and particle repetitions. This effect was particularly pronounced in geometric surface variations and distinct inhomogeneity boundaries. Discrepancies and limitations with the current MC modelling ability were discovered by implementing identical shared components when developing models. A design flaw was discovered with the BEAMnrc MLCE module where particle collimation does not appropriately collimate the beam throughout the module. A solution was implemented and reduced BEAMnrc model output errors from 5% to 2% without the use of any correction factors. The APPLICAT module was determined to not sufficiently model applicators to the required standard. The inclusion of applicators into the beam models resulted in underestimation of dose delivered up to 3% and was proportional to applicator size. The applicator misrepresentation used in Monaco 5 and BEAMnrc models results in dose prediction errors for the 6 cm × 6 cm and 10 cm × 10 cm applicators, increasing in magnitude with increasing distance from 100 cm SSD. The 6 cm × 6 cm applicator was found to be most susceptible resulting in output error predictions of -2% to 4% over SSDs 97 to 105 cm respectively. To achieve the level of accuracy required to be considered a gold standard, MC models require accurate representation of the linear accelerator, appropriate incident spectrum and refinements in modelling. The independent beam model development process has proven vital in determining limitations in current electron beam modelling that would otherwise remain undetected. Currently, Elekta electron MC models do not meet the gold standard as correction factors are required to correct for absolute dosimetry predictions. If sufficient effort and refinements are made, in-house MC models can produce superior relative dosimetry results compared to commercial TPSs due the greater control over the end-to-end dose prediction process with the disadvantage of increased calculation time. An in-house, independent electron MC dose calculation toolkit has proven a valuable tool in validating the performance and understanding the limitations of Elekta’s electron MC TPS, Monaco 5.