Purpose:Electronic portal imaging devices (EPIDs) are increasingly used for IMRT dose verification, both pretreatment andin vivo. In this study, an earlier developed backprojection model has been modified to avoid the need for patient‐specific transmission measurements and, consequently, leads to a faster procedure.Methods:Currently, the transmission, an essential ingredient of the backprojection model, is estimated from the ratio of EPID measurements with and without a phantom/patient in the beam. Thus, an additional irradiation to obtain “open images” under the same conditions as the actual phantom/patient irradiation is required. However, by calculating the transmission of the phantom/patient in the direction of the beam instead of using open images, this extra measurement can be avoided. This was achieved by using a model that includes the effect of beam hardening and off‐axis dependence of the EPID response on photon beam spectral changes. The parameters in the model were empirically obtained by performing EPID measurements using polystyrene slab phantoms of different thickness in 6, 10, and 18 MV photon beams. A theoretical analysis to verify the sensitivity of the model with patient thickness changes was performed. The new model was finally applied for the analysis of EPID dose verification measurements of step‐and‐shoot IMRT treatments of head and neck, lung, breast, cervix, prostate, and rectum patients. All measurements were carried out using Elekta SL20i linear accelerators equipped with a hydrogenated amorphous silicon EPID, and the IMRT plans were made usingPINNACLE software (Philips Medical Systems).Results:The results showed generally good agreement with the dose determined using the old model applying the measured transmission. The average differences between EPID‐basedin vivo dose at the isocenter determined using either the new model for transmission and its measured value were , , and for 47 patients treated with 6, 10, and 18 MV IMRT beams, respectively. For the same group of patients, the differences in mean analysis (3% maximum dose, 3 mm) were , , and , respectively. For a subgroup of 11 patients, pretreatment verification was also performed, showing similar dose differences at the isocenter: , , and , with somewhat lower mean difference values: , , and , respectively. Clinical implementation of the new model would save 450 h/yr spent in measurement of open images.Conclusions:It can be concluded that calculating instead of measuring the transmission leads to differences in the isocenter dose generally smaller than 2% (2.6% for 6 MV photon beams forin vivo dose) and yielded only slightly higher ‐evaluation parameter values in planes through the isocenter. Hence, the new model is suitable for clinical implementation and measurement of open images can be omitted.