Patient-specific models of the cerebral aneurysms and connected vessels were constructed from the 3DRA images using seeded region growing algorithms that reconstruct the vascular topology, followed by iso-surface deformable models that recover the vascular geometry (18 , 19 ). As much of the proximal parent artery visible in the 3D images was included in the models to ensure proper representation of secondary flows and inflow to the aneurysms (20 (link)). Unstructured tetrahedral grids were then generated with a resolution of 0.01 to 0.02 cm for CFD simulation. Pulsatile blood flow simulations were carried out by numerically solving the 3D Navier-Stokes equations for a Newtonian incompressible fluid under the assumption of rigid vessel walls. Because patient-specific flow conditions were not available typical flow waveforms derived from measurements on normal subjects at different heart rates were used to prescribe inlet boundary conditions (21 (link), 22 ). The flow waveforms were scaled to achieve a given mean wall shear stress (WSS) at the inlets, which were located in internal carotid, vertebral or basilar arteries. A total of 5 simulations were carried out for each aneurysm, two under pulsatile conditions corresponding to heart rates of 60 and 100 bpm and a mean inlet wall shear stress of 15 dyn/cm2, and three under steady flow conditions named low, medium and high flow rates corresponding to inlet WSS of 10, 15 and 20 dyn/cm2, respectively. The unsteady flow solutions were advanced in time using 100 timesteps per cardiac cycle for two cycles using a fully implicit scheme and efficient solution algorithms (23 , 24 (link)). Results of the second cycle were used for hemodynamic aneurysm characterization.