Purpose: Recent advances in 3D-PCMRI (phase contrast MRI) sequences allow for measuring the complex hemodynamics in cerebral arteries. However, the small size of these vessels vs spatial resolution can lead to non-negligible partial volume artifacts, which must be taken into account when computing blood flow rates. For this purpose, we combined the velocity information provided by 3D-PCMRI with vessel geometry measured with 3DTOF (time of flight MRI) or 3DRA (3D rotational angiography) to correct the partial volume effects in flow rate assessments. Methods: The proposed methodology was first tested in vitro on cylindrical and patient specific vessels subject to fully controlled pulsatile flows. Both 2D- and 3D-PCMRI measurements using various spatial resolutions ranging from 20 to 1.3 voxels per vessel diameter were analyzed and compared with flowmeter baseline. Second, 3DTOF, 2D- and 3D-PCMRI measurements were performed in vivo on 35 patients harboring internal carotid artery (ICA) aneurysms indicated for endovascular treatments requiring 3DRA imaging. Results: The in vitro 2D- and 3D-PCMRI mean flow rates assessed with partial volume correction showed very low sensitivity to the acquisition resolution above ≈ voxels per vessel diameter while uncorrected flow rates deviated critically when decreasing the spatial resolution. 3D-PCMRI flow rates measured in vivo in ICA agreed very well with 2D-PCMRI data and a good flow conservation was observed at the C7 bifurcation. Globally, partial volume correction led to 10–15% lower flow rates than uncorrected values as those reported in most of the published studies on intracranial flows. Conclusion: Partial volume correction may improve the accuracy of PCMRI flow rate measurements especially in small vessels such as intracranial arteries. Magn Reson Med 79:129–140, 2018.