In western populations, cerebral aneurysms develop in approximately 4% of humans and they involve the risk of rupture. Blood flow patterns are of interest for understanding the pathogenesis of the lesions and may eventually contribute to deciding on the most efficient treatment procedure for a specific patient. Velocity mapping with phase-contrast magnetic resonance angiography (PC-MRA) is a non-invasive method for performing in vivo measurements on blood velocity. Several hemodynamic properties can either be derived directly from these measurements or a flow field with all its parameters can be simulated on the basis of the measurements. For both approaches, the accuracy of the PC-MRA data and subsequent modeling must be validated. Therefore, a realistic transient flow field in a well-defined patient-specific silicone phantom was investigated. Velocity investigations with PC-MRA in a 3 Tesla MR scanner, laser Doppler velocimetry (LDV) and computational fluid dynamics (CFD) were performed in the same model under equal flow conditions and compared to each other. The results showed that PC-MRA was qualitatively similar to LDV and CFD, but showed notable quantitative differences, while LDV and CFD agreed well. The accuracy of velocity quantification by PC-MRA was best in straight artery regions with the measurement plane being perpendicular to the primary flow direction. The accuracy decreased in regions with disturbed flow and in cases where the measurement plane was not perpendicular to the primary flow. Due to these findings, it is appropriate to use PC-MRA as the inlet and outlet conditions for numerical simulations to calculate velocities and shear stresses in disturbed regions like aneurysms, rather than derive these values directly from the full PC-MRA measured velocity field. (c) 2009 John Wiley & Sons, Ltd.