Segmentation et quantification des couches rétiniennes dans des images de tomographie de cohérence optique, dans le cas de sujets sains et pathologiques

Abstract : Optical coherence tomography (OCT) is a non-invasive imaging technique, based on the principle of interferometry. Thus, OCT is now a standard examination for the detection and the monitoring of retinal diseases including macular degeneration. In this context, the first objective of this thesis is to propose a new method for the segmentation of OCT images of healthy subjects. The proposed method exploits prior knowledge on the structure and the appearearance of the retinal layers. It is based on a combination of local and global segmentation algorithms, including active contours, k-means and Markov random fields. Thus, eight retinal layers can be detected, including the inner segments (IS) of photoreceptors. However, the slow evolution of this disease makes the evaluation of these therapies difficult. The second objective of this thesis is then to extend the scope of the method developed for healthy subjects to retinitis pigmentosa subjects. We have developed a new parametric deformable model that incorporates a priori information by adding a constraint of approximate parallelism, which is more robust in the presence of pathologies. In both healthy and pathological study cases, we performed a comprehensive qualitative and quantitative assessrnent of the proposed methods. We evaluated the accuracv of the segmentation of interfaces between layers, and, in the case of healthy subjects, the accuracy of the segmentation of interfaces between layers, and, in the case of healthy subjects, the precision of thickness measurements derived from the segmentation. This study was conducted on a large image database. These evaluations show a very good agreement anda strong correlation between automatic segmentation and segmentation done manually by an expert.
Mots-clés : Oeil Cohérence optique
Complete list of metadatas

Cited literature [75 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/pastel-00719456
Contributor : Abes Star <>
Submitted on : Thursday, July 19, 2012 - 6:17:29 PM
Last modification on : Thursday, October 17, 2019 - 12:36:06 PM
Long-term archiving on : Saturday, October 20, 2012 - 3:20:09 AM

File

these_Ghorbel.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : pastel-00719456, version 1

Collections

Citation

Itebeddine Ghorbel. Segmentation et quantification des couches rétiniennes dans des images de tomographie de cohérence optique, dans le cas de sujets sains et pathologiques. Autre. Télécom ParisTech, 2012. Français. ⟨NNT : 2012ENST0013⟩. ⟨pastel-00719456⟩

Share

Metrics

Record views

978

Files downloads

504