Efficient Monte Carlo simulation of multileaf collimators using geometry-related variance-reduction techniques.

A technique for accelerating the simulation of multileaf collimators with Monte Carlo methods is presented. This technique, which will be referred to as the movable-skin method, is based on geometrical modifications that do not alter the physical shape of the leaves, but that affect the logical way in which the Monte Carlo code processes the geometry. Zones of the geometry from which secondary radiation can emerge are defined as skins and the radiation transport throughout these zones is simulated accurately, while transport in non-skin zones is modelled approximately. The skins method is general and can be applied to most of the radiation transport Monte Carlo codes used in radiotherapy. The code AUTOLINAC for the automatic generation of the geometry file and the physical parameters required in a simulation of a linac with the Monte Carlo code PENELOPE is also introduced. This code has a modularized library of all Varian Clinac machines with their multileaf collimators and electron applicators. AUTOLINAC automatically determines the position of skins and the parameter values employed for other variance-reduction techniques that are adequate for the simulation of a linac. Using the adaptive variance-reduction techniques presented here it is possible to simulate with PENELOPE an entire linac with a fully closed multileaf collimator in two hours. For this benchmark a single core of a 2.8 GHz processor was used and 2% statistical uncertainty (1sigma) of the absorbed dose in water was reached with a voxel size of 2 x 2 x 2 mm(3). Several configurations of the multileaf collimator were simulated and the results were found to be in excellent agreement with experimental measurements.