The purpose of the study was to investigate whether visual inspection of maximum intensity projection (MIP) images is a reliable method for assessment of the severity of renal artery stenoses (RAS). Therefore, 20 RAS were investigated with helical CT. Native axial images and MIP images were analysed separately or in combination ("two-step method"). A phantom study was performed to investigate the influence of window setting on apparent stenosis severity. Accuracy for diagnosis of 50-69% and 70-99% RAS and the mean inter-observer agreement were 82.5%, 77.5% and 82.5% respectively, using study of MIP images alone, and 100%, 97.5% and 95% using the "two-step method". The phantom study showed that the apparent severity of vascular stenoses on MIP images depends on the selected window centre values. It is concluded that combined evaluation of both MIP images and native axial images is more reliable than study of MIP images alone for assessing the severity of RAS.