Model Based Multiscale Detection and Reconstruction of 3D Vessels - Inria - Institut national de recherche en sciences et technologies du numérique Access content directly
Reports Year : 1998

Model Based Multiscale Detection and Reconstruction of 3D Vessels

Grégoire Malandain
Nicholas Ayache
  • Function : Author
  • PersonId : 833436

Abstract

The segmentation of 3D brain vessels is an important issue for physicians in order to operate an aneurysm. We introduce new vessel models for selecting a subset of interesting points near the vessel center. We also present a new approach to segment and reconstruct 3D brain vessels. The response at one scale is obtained by integrating along a circle the first derivative of the intensity in the radial direction. We also use a vessel model to choose a good parameter for a gamma-normalizatio- n of the response obtained at each scale. Once the parameter gamma is fixed, we find the relation between a vessel radius and the scale at which it is detected. From the multiscale response, we create a smoothed skeleton of the vessels and we reconstruct the vessels from their centerlines and their radii. The method has been tested on a large variety of 3D images of cerebral vessels, with excellent results. Vessels of various size and contrast are detected with remarkable robustness, even when they are close or tangent to another vessel, and most junctions are preserved. Results are obtained in a few minutes on a Dec-Alpha workstation, for a 128^3 image. This work was done in collaboration with General Electric Medical Systems Europe (GEMSE).

Domains

Other [cs.OH]
Fichier principal
Vignette du fichier
RR-3442.pdf (1.06 Mo) Télécharger le fichier

Dates and versions

inria-00073248 , version 1 (24-05-2006)

Identifiers

  • HAL Id : inria-00073248 , version 1

Cite

Karl Krissian, Grégoire Malandain, Nicholas Ayache. Model Based Multiscale Detection and Reconstruction of 3D Vessels. RR-3442, INRIA. 1998. ⟨inria-00073248⟩
113 View
265 Download

Share

Gmail Facebook Twitter LinkedIn More