We discuss in this paper the problem of localizing and quantifying intracranial aneurysms. Assuming that the segmentation of medical images is done, and that a 3D representation of the vascular tree is available, we present a new automatic algorithm to extract vessels centerlines. Aneurysms are then automatically detected by studying variations of vessels diameters. Once an aneurysm is detected, we give measures that are important to decide its treatment. The name of the aneurysm-carrying vessel is computed using an inexact graph matching technique. The proposed approach is evaluated on segmented real images issued from Magnetic Resonance Angiography (MRA) and CT scan.