Computational fluid dynamics (CFD) is a technique that provides valuable information on blood flow from the vascular geometry, helping to understand, diagnose and predict the outcome of vascular diseases. However, the resolution of current medical images is unsatisfactory, and it is still difficult to extractblood vessels, especially those with large and complex geometries such as cerebral artery networks. In this thesis, we proposed a two-step framework to produceCFD-ready meshes from a simplified representation of vascular networks by their centerlines. In the first stage, to address the shortcomings of the centerline-based representation (scattered, noisy), a modeling step was introduced to reconstruct an anatomically realistic model from a-priori knowledge on the vessels and bifurcations geometries. Next, a meshing step was developed to create a high-quality volume mesh with structured, flow- oriented hexahedral cells that meets the requirements of CFD. Using this software, we have created an open data set of meshes for CFD of whole cerebral artery networks that can be used for evaluation of medical devices and hemodynamic studies. This software contributes to solve the shortcomings of current meshing methods and enable the construction of a large-scale CFD database of cerebral arterial networks.