Extraction et caractérisation de régions saines et pathologiques à partir de micro-tomographie RX du système vasculaire cérébral

Xavier Descombes (1) , Ahmed Gamal Eldin (1) , Franck Plouraboué (2) , Caroline Fonta (3) , Raphaël Serduc (4) , Géraldine Le Duc (5) , Timm Weitkamp (5)

1 ARIANA - Inverse problems in earth monitoring
2 IMFT - Institut de mécanique des fluides de Toulouse
3 CERCO - Centre de recherche cerveau et cognition
4 UJF - Université Joseph Fourier - Grenoble 1
5 ESRF - European Synchrotron Radiation Facility

In this paper, we consider X-ray micro-tomography representing the brain vascular network. We define the local vascular territories as the regions obtained after a watershed algorithm applied on the distance map. The obtained graph is then regularized by a Markov random field approach. The optimization is performed using a graph cut algorithm. We show that the resulting segmentation exhibits three classes corresponding to normal tissue, tumour and an intermediate region.

Domaines: Traitement des images [eess.IV]

Xavier Descombes, Ahmed Gamal Eldin, Franck Plouraboué, Caroline Fonta, Raphaël Serduc, et al.. Extraction et caractérisation de régions saines et pathologiques à partir de micro-tomographie RX du système vasculaire cérébral. GRETSI, Sep 2011, Bordeaux, France. ⟨inria-00625525⟩