Optimisation of variance reduction techniques in EGSnrc Monte Carlo for a 6 MV photon beam of an Elekta Synergy linear accelerator

Objective: Monte Carlo (MC) simulations are considered to be the most accurate form of algorithm for dose calculation. However, the main obstacle to using MC, especially in clinical routine, is the simulation time needed to gain results below a desirable level of uncertainty. Variance reduction techniques (VRTs) have been introduced to reduce the simulation time while maintaining the uncertainty at an acceptable level. The aim of this study is to investigate and optimize the VRTs implemented in EGSnrc MC code, BEAMnrc and DOSXYZnrc. Methodology: The BEAMnrc user code was used to model a 10 x 10 cm(2) field size of a 6 MV photon beam from an Elekta Synergy linear accelerator. The DOSXYZnrc user code was used to model a water phantom. The effects of different VRTs on the simulation efficiency were investigated either individually or in combination. The directional bremsstrahlung splitting (DBS) technique was investigated further to find the optimum splitting number and splitting field radius. For DOSXYZnrc, the photon splitting was investigated to find the best combination with the VRTs in BEAMnrc and to find the optimum splitting number. Finally, the best combination of VRTs in both BEAMnrc and DOSXYZnrc was compared with the corresponding phase space (PHSP) simulation source. Results: The DBS technique was found to be the most efficient. The optimum splitting number was found to be 10,000 and 15,000 with and without electron splitting, respectively. For the DBS splitting field radius, overestimating by up to 3 cm would be sufficient without causing a significant loss in efficiency. For both BEAMnrc and DOSXYZnrc, the combination of DBS, bremsstrahlung cross-section enhancement, range rejection with 2 MeV and photon splitting (with optimum splitting number of 35) was the most efficient, and was about 8% less efficient than PHSP simulation. Conclusion: The VRTs implemented in EGSnrc MC code made it possible to achieve an acceptably small uncertainty within a reasonable simulation time, if optimised properly. The combination of VRTs presented in this study eliminates the need to spare a large amount of disk space, and where parallel computing could allow for MC dose calculation in real-time adaptive treatment planning. (C) 2021 The Author(s). Published by Elsevier B.V. on behalf of King Saud University.