The purpose of this study was to demonstrate the feasibility of using a custom gradient sequence on an unmodified 3T magnetic resonance imaging (MRI) scanner to perform magnetic resonance navigation (MRN) by investigating the blood flow control method invivo, reproducing the obtained rheology in a phantom mimicking porcine hepatic arterial anatomy, injecting magnetized microbead aggregates through an implantable catheter, and steering the aggregates across arterial bifurcations for selective tumor embolization. In the first phase, arterial hepatic velocity was measured using cine phase-contrast imaging in seven pigs under free-flow conditions and controlled-flow conditions, whereby a balloon catheter is used to occlude arterial flow and saline is injected at different rates. Three of the seven pigs previously underwent selective lobe embolization to simulate a chemoembolization procedure. In the second phase, the measured invivo controlled-flow velocities were approximately reproduced in a Y-shaped vascular bifurcation phantom by injecting saline at an average rate of 0.6mL/s with a pulsatile component. Aggregates of 200-μm magnetized particles were steered toward the right or left hepatic branch using a 20-mT/m MRN gradient. The phantom was oriented at 0°, 45°, and 90° with respect to the B0 magnetic field. The steering differences between left-right gradient and baseline were calculated using Fisher's exact test. A theoretical model of the trajectory of the aggregate within the main phantom branch taking into account gravity, magnetic force, and hydrodynamic drag was also designed, solved, and validated against the experimental results to characterize the physical limitations of the method. At an injection rate of 0.5mL/s, the average flow velocity decreased from 20±15 to 8.4±5.0cm/s after occlusion in nonembolized pigs and from 13.6±2.0 to 5.4±3.0cm/s in previously embolized pigs. The pulsatility index measured to be 1.7±1.8 and 1.1±0.1 for nonembolized and embolized pigs, respectively, decreased to 0.6±0.4 and 0.7±0.3 after occlusion. For MRN performed at each orientation, the left-right distribution of aggregates was 55%, 25%, and 75% on baseline and 100%, 100%, and 100% (P<0.001, P=0.003, P=0.003) after the application of MRN, respectively. According to the theoretical model, the aggregate reaches a stable transverse position located toward the direction of the gradient at a distance equal to 5.8% of the radius away from the centerline within 0.11s, at which point the aggregate will have transited through a longitudinal distance of 1.0mm from its release position. In this study, we showed that the use of a balloon catheter reduces arterial hepatic flow magnitude and variation with the aim to reduce steering failures caused by fast blood flow rates and low magnetic steering forces. A mathematical model confirmed that the reduced flow rate is low enough to maximize steering ratio. After reproducing the flow rate in a vascular bifurcation phantom, we demonstrated the feasibility of MRN after injection of microparticle aggregates through a dedicated injector. This work is an important step leading to MRN-based selective embolization techniques in humans.