Skip to Main content Skip to Navigation
Theses

Contribution à la segmentation des réseaux vasculaires cérébraux obtenus en IRM. Intégration de connaissance anatomique pour le guidage d'outils de morphologie mathématique

Résumé : Analysing cerebral MRA (magnetic resonance angiography) is a hard task for radiologists, because of the large size of the data and the increasing number of exams being performed. Creation of cerebral vessel segmentation methods from such images then constitutes a research area of great importance in medical imaging. This thesis is devoted to the development of such methods. It is especially focused on their ability to adapt their behaviour to the processed images and their semantic value. This concept of adaptivity is developed by considering high level anatomical knowledge which can be used for guidance of image processing tools. The first part of the presented work consists in proposing preliminary solutions for knowledge modelling. The atlas notion, which has already been successfully used for non vascular structures, is then developed. Two kinds of atlas are proposed, each one taking advantage of multi-modality (angiographic and morphologic) properties of the considered images in order to model anatomical knowledge elements related to brain vessels. The second part of the work deals with the development of segmentation methods using this knowledge for guiding mathematical morphology tools. These methods, based on region-growing, watershed, grey-level hit-or-miss transform and homotopic thinning, use the proposed atlases to fit or constraint the behaviour of these image processing tools with respect to the image properties. This thesis can be considered as an introduction to a new methodology of vascular structure segmentation, which tends to fuse the potential of the existing image processing tools with learning and knowledge based strategies which are generally only used by human specialists.
Complete list of metadatas

Cited literature [192 references]  Display  Hide  Download

https://hal.univ-reims.fr/tel-01695497
Contributor : Nicolas Passat <>
Submitted on : Monday, January 29, 2018 - 2:05:34 PM
Last modification on : Thursday, April 23, 2020 - 2:26:35 PM
Document(s) archivé(s) le : Friday, May 25, 2018 - 12:22:38 PM

File

Passat PhD 2005.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-01695497, version 1

Collections

Citation

Nicolas Passat. Contribution à la segmentation des réseaux vasculaires cérébraux obtenus en IRM. Intégration de connaissance anatomique pour le guidage d'outils de morphologie mathématique. Traitement des images [eess.IV]. Université Strasbourg 1, 2005. Français. ⟨tel-01695497⟩

Share

Metrics

Record views

87

Files downloads

392