In recent years, EPIDs have been used for pre-treatment IMRT verification. Although EPID lag and signal nonlinearities have been investigated, they have not been implemented in the verification process. In dynamic sliding-window IMRT delivery, the dose delivered, and the time between the end of dose delivery and the end of image acquisition differ between pixels. The resulting differences in lag and signal-response across the image can cause artificial asymmetries and amplitude changes in measured EPID dose images. These artifacts alter the agreement between measured and predicted images, potentially complicating the assessment of clinical IMRT verifications. A method of 2-D (pixel-by-pixel) correction was developed based on data from sets of experiments performed to independently quantify the lag and nonlinearity characteristics of Varian's aS500 EPID. To test the correction, it was applied to two sweeping window 10×10 cm2 fields that differ only in sweeping direction. The correction resolved discrepancies in the symmetry between these two cases, and the differences between measured and predicted amplitudes evident when small numbers of MUs were delivered. To illustrate its potential use, the correction technique was applied to a measured image of a clinical IMRT field that produced a relatively poor verification result. The correction partially accounted for discrepancies between measured and Eclipse-predicted images of this field, reducing the percentage of pixels failing a Gamma analysis (3 %, 3 mm) from 8.5 to 5.6 %. This correction technique can be used to help resolve the source of discrepancies in troublesome clinical IMRT verifications.