Manifold-enhanced Segmentation through Random Walks on Linear Subspace Priors

Pierre-Yves Baudin (1, 2) , Noura Azzabou (3, 4, 5) , Pierre G. Carlier (3, 4, 5) , Nikos Paragios (1, 6, 7, 8, 9, 10)

1 GALEN - Organ Modeling through Extraction, Representation and Understanding of Medical Image Content
2 CVN - Centre de vision numérique
3 Institut de Myologie
4 MIRCEN - Molecular Imaging Research Center [Fontenay-aux-Roses]
5 UPMC - Université Pierre et Marie Curie - Paris 6
6 imagine [Marne-la-Vallée]
7 LIGM - Laboratoire d'Informatique Gaspard-Monge
8 UPEM - Université Paris-Est Marne-la-Vallée
9 ENPC - École des Ponts ParisTech
10 MAS - Mathématiques Appliquées aux Systèmes - EA 4037

In this paper we propose a novel method for knowledge-based segmentation. Our contribution lies on the introduction of linear sub-spaces constraints within the random-walk segmentation framework. Prior knowledge is obtained through principal component analysis that is then combined with conventional boundary constraints for image segmentation. The approach is validated on a challenging clinical setting that is multicomponent segmentation of the human upper leg skeletal muscle in Magnetic Resonance Imaging, where there is limited visual differentiation support between muscle classes.

Domaines: Vision par ordinateur et reconnaissance de formes [cs.CV]; Traitement des images [eess.IV]

Pierre-Yves Baudin, Noura Azzabou, Pierre G. Carlier, Nikos Paragios. Manifold-enhanced Segmentation through Random Walks on Linear Subspace Priors. British Machine Vision Conference, 2012, United Kingdom. pp.51.1-51.10. ⟨hal-00773635⟩