The nonwater-equivalent energy response of electronic portal imaging devices (EPIDs) is a major obstacle to using them for linear accelerator (linac) beam parameter verification. In this study, we propose an EPID-based machine quality assurance (QA) application that uses a model-based radiation transport algorithm to convert EPID-measured images into water-equivalent dose distributions that can be used to assess beam flatness and symmetry. An in-house developed, model-based radiation transport algorithm was used to estimate the incident beam fluence from measured EPID images and convert it into either 3D dose distributions in a virtual water tank or 2D water-equivalent dose distributions in a virtual ion chamber array. The conversion algorithm was validated using independent measurements in a scanning water tank and a reference ion chamber array under symmetric and also intentionally detuned (i.e., asymmetric) beam conditions. For symmetric fields, EPID-reconstructed percentage depth dose distributions (PDDs) agreed with water tank measurements to within 1% beyond the first 10 mm of depth. Beam profile comparisons showed differences within 1% in low dose-gradient regions. For all symmetric and intentionally asymmetric fields, beam flatness and symmetry derived from reconstructed images agreed with reference measurements to within 0.2% and 0.3%, respectively. The model demonstrated high sensitivity to the controlled beam asymmetries and steering distortions, with EPID-reconstructed metrics closely matching reference water-equivalent measurements and significantly outperforming metrics derived from raw EPID images. The proposed model-based algorithm enables accurate conversion of EPID images into water-equivalent dose distributions, facilitating accurate determination of beam flatness and symmetry. This application addresses some limitations of the previously proposed EPID-based linac QA techniques, which are limited to nonwater-equivalent constancy checks, and supports the use of EPIDs as robust dosimetry tools for linac radiation beam parameter verification.
Kutuzov et al. (Fri,) studied this question.