15. EPID-based pretreatment quality assurance: Dosimetric evaluation of application software

Abstract

Introduction With the prevalence of gantry-fixed and rotational IMRT, pretreatment dosimetric Quality Assurance (QA) is a good practice. The pretreatment control is a dosimetric measurement under the treatment device which allows the virtual information transmitted from the Treatment Planning System (TPS) to be linked to the actual dose delivery. As known for a few years, the Electronic Portal Imaging Device (EPID) could potentially be used for dosimetric and routine machine QA purposes, other than patient setup verification which is its intended initial purpose. Based on EPID acquisitions, software dedicated to patient specific QA allow comparing, in dosimetric terms, the portal image acquired under the treatment machine with a predicted image computed under ideal conditions. In this work, we evaluate an independent and automatic solution integrating both an absolute dose conversion model of the portal images, and a prediction model computing theoretical 2D dose distributions from the TPS Dicom RT Plan file. The purpose of the study is to describe the relevance of the prediction model, with reference to TPS calculations, and the conversion model compared to a true dose measurement. Methods Predictive 2D dose images and dose converted portal images are computed with the ThinkQA/EpiBeam solution (Dosisoft S.A.). By convention, these two images are expressed in absolute dose at isocentre level under 5 cm depth of water. Reference dose calculations and dose measurements are performed, in homogenous phantom in the same conventional conditions, respectively by the TPS Isogray (Dosisoft S.A.) or Monaco (Elekta) and with the Matrixx (IBA). EPID images are acquired with the iViewGT system (Elekta). The evaluation is carried out with a 6 MV photon beam delivered by an Elekta Synergy treatment unit and includes static beams, IMRT and VMAT plans. Dosimetric comparisons are performed according to 2D gamma-index agreement (GAI) values. Results Similar dose distributions are obtained with the prediction model and both TPS. GAI are greater than 95% with restrictive gamma-index criteria (2% dose, 2 mm distanceto-agreement) for dose region receiving more than 10% of the maximum dose. Conversion model returns pertinent experimental dose values compared to the Matrixx measurements. GAI of about 90% (3%, 3 mm) or 95% (3%, 4 mm) are achievable according to the poor spatial resolution (7 mm) of the Matrixx device. Conclusions The pretreatment software solution studied returns accurate predictive and converted dose images and therefore, a relevant comparison between TPS dose calculation and beam delivery. Automatic computation of the predictive image by the system frees the users from the need of creating specific QA plans in the TPS.