The purpose of this study was to develop, implement and validate a method for portal dosimetry with an amorphous silicon EPID for a wide energy range. Analytic functions were applied in order to correct for nonlinearities in detector response with dose rate, irradiation time and total dose. EPID scattering processes were corrected for by means of empirically determined convolution kernels. For a variety of rectangular and irregularly shaped fields, head scatter factors determined from central axis portal dose values and those measured with an ionization chamber showed a maximum deviation of 0.5%. The accuracy of our method was further investigated for pretreatment IMRT verification (i.e. without absorbers in the beam). The agreement between EPID and film dosimetry was quantified using gamma (gamma) evaluation, with 2% dose and 2 mm distance-to-agreement criteria. All gamma-distributions showed a gamma(mean) < 0.5, a 99th percentile <1.5 and a fraction of pixels with gamma > 1 smaller than 7%. The number of monitor units delivered by single segments of the IMRT fields could be extracted from the portal images with high accuracy. Measured and delivered doses were within +/-3% for more than 98% of data points. Ghosting effects were found to have limited effects on dosimetric IMRT verification.