Résumé : Detection of tubular structures in 3D images is an important issue for vascular detection in medical imaging. We present in this paper a new approach for centerline detection and reconstruction of 3D tubular structures. Several models of vessels are introduced for estimating the sensivity of the image second order derivatives according to elliptical cross-section, to curvature of the axis, or to partial volume effects. Our approach uses a multiscale analysis for extracting vessels of different sizes according to the scale. For a given model of vessel, we derive an analytic expression of the relationship between the radius of the structure and the scale at which it is detected. The algorithm gives both centerline extraction and radius estimation of the vessels allowing their reconstruction. The method has been tested on both synthetic and real images, with encouraging results. This work was done in collaboration with GEMS (General Electric Medical Systems, Buc, France).