Inclusion of the treatment couch in portal dose image prediction for high precision EPID dosimetry
Purpose: When comparing predicted portal dose images (PDIs) to PDIs acquired by an EPID during treatment delivery, differences are often observed. These differences may be partially attributed to beam attenuation by parts of the treatment couch not taken into account in the PDI prediction. In order to improve the agreement, a model for the treatment couch was derived and included in the PDI prediction. Methods: A CT scan was used to model the couch top. The model for the couch top base was derived by iteratively matching the predicted and measured PDIs for gantry angles of 0°, 45°, and 90°. For PDI prediction, the modeled treatment couch was added to the CT scan of a patient or phantom by using the recorded couch positions from the record and verify system. To validate the couch model, PDI measurements were performed for a range of couch positions and gantry angles, both with and without an anatomical phantom in the beam. Results: After including the couch model in the PDI prediction for beams passing through the couch without phantom, the mean local dose differences between measured and predicted PDIs were reduced from up to 5.5% to less than 1.0% at each gantry angle. Similar results were obtained for measurements with a lung phantom on the couch. Although the couch model was originally derived by using a 6 MV photon beam, the results showed that it is also applicable for a 10 MV beam. Conclusions: A model of the treatment couch was derived and included in the PDI prediction, yielding a substantially improved agreement between measured and predicted PDIs, which makes interpretation of the observed deviations more straightforward.