The fluid dynamic performance of mechanical heart valves differs from normal valves and thus is considered related to late clinical complications in patients. Since flow patterns evolving around heart valves are complex in space and time, flow visualization based on time-resolved 3D velocity data might add important information regarding the performance of specific valve designs in vivo. However, previous cine 3D techniques for three-directional phase-contrast velocity mapping suffer from long scan duration and therefore might hamper assessment in patients. A hybrid 3D phase-contrast sequence combining segmented k-space acquisition with short EPI readout trains is presented with its validation in vitro. The technique was applied to study flow patterns downstream from a bileaflet aortic prosthesis in six patients. Navigator echoes were incorporated for respiratory motion compensation. Before flow visualization, spurious phase errors due to concomitant gradient fields and eddy currents were corrected. Flow visualization was based on particle paths and animated velocity vector plots. Dedicated algorithms for particle path integration were implemented to account for the considerable motion of the ascending aorta during the cardiac cycle. A distinct flow pattern reflecting the valve design was observed closest to the valve during early flow acceleration. Reverse flow occurred adjacent to high velocity jets and above the hinge housings. Later in systole, flow became confined to the central vessel area and reverse flow along the inner aortic curvature developed. Further downstream from the valve, flow patterns varied considerably among patients, indicating the impact of varying aortic anatomy in vivo. It is concluded that MR velocity mapping is a potential tool for studying 3D flow patterns evolving around heart valve prostheses in humans. J. Magn. Reson. Imaging 2001;13:690-698. Copyright 2001 Wiley-Liss, Inc.