Experimental validation of a Monte Carlo model to predict EPID images for online verification in radiotherapy

One proposed method this past decade for online verification of dose delivery in radiotherapy consists in comparing images acquired using Electronic Portal Imaging Devices (EPIDs) with predicted dose images computed using Monte Carlo simulations. The objective of this study is twofold: first, to demonstrate that the Monte Carlo simulation code PENELOPE is suitable to compute reliably portal images, and second, to validate a portal prediction model against measurements. Portal images were acquired with a fluoroscopic EPID and a Saturne 43 accelerator (12 MV photons) for different field sizes, both with and without a 30×30×30 cm³ water phantom in the beam. Monte Carlo simulations of the accelerator and the EPID were performed using the Monte Carlo simulation code PENELOPE. Several EPID models were tested, differing by the level of complexity in the geometry. This study shows that the EPID signal can be realistically predicted by a simple three-layer model, including two layers to describe the fluorescent screen and an additional water layer to take into account optical photons backscattering within the EPID structure. Using this model, 2D gamma index values less than 1 were obtained for 96% of the pixels for fields of 10×10 cm² or less and for 90% of the pixels for larger field sizes. These results exhibit that the EPID model determined in this study allows computing accurately portal images of open fields, with or without an object in the beam.