Optimizing portal dose calculation for an amorphous silicon detector using Swiss Monte Carlo plan

Abstract

Purpose: Modern treatment planning systems (TPS) are able to calculate doses within the patient for numerous delivery techniques as e. g. intensity modulated radiation therapy (IMRT). Even dose predictions to an electronic portal image device (EPID) are available in some TPS, but with limitations in accuracy. With the steadily increasing number of facilities using EPIDs for pre-treatment and treatment verification, the desire of calculating accurate EPID dose distributions is growing. A solution for this problem is the use of Monte Carlo (MC) methods. Aims of this study were firstly to implement geometries of an amorphous silicon based EPID with varying levels of geometry complexity. Secondly to analyze the differences between simulation results and measurements for each geometry. Thirdly, to compare different transport algorithms within all EPID geometries in a flexible C++ MC environment. Materials and Methods: In this work three geometry sets, representing the EPID, are implemented and investigated. To gain flexibility in the MC environment geometry and particle transport code are independent. That allows the user to select between the transport algorithms EGSnrc, VMC++ and PIN (an in-house developed transport code) while using one of the implemented geometries of the EPID. For all implemented EPID geometries dose distributions were calculated for 6 MV and 15 MV beams using different transport algorithms and are then compared with measurements. Results: A very simple geometry, consisting of a water slab, is not capable to reproduce measurements, whereas 8 material layers perform well. The more layers with different materials are used, the longer last the calculations. EGSnrc and VMC++ lead to dosimetrically equal results. Gamma analysis between calculated and measured EPID dose distributions, using a dose difference criterion of 3% and a distance to agreement criterion of 3 mm, revealed a gamma value < 1 within more than 95% of all pixels, that have a higher dose than 3% of the measured maximum dose. Conclusions: The same geometry can be used to compare transport codes within the flexible C++ MC environment. A simplified 8 layer geometry results in similar EPID dose distributions compared with the accurate 24 layer geometry by gaining about a factor of 6 in CPU-time. The implementation of the amorphous silicon detector in our MC system has the potential to perform independent pre-treatment as well as treatment verification. © 2007 IOP Publishing Ltd.