Nouvelles méthodes de segmentation en imagerie tomographique volumique à faisceau conique dentaire

Abstract : Cone-Beam computed tomography (CBCT) is the new standard imaging method for dental practitioners. The image processing field of CBCT data is still underdeveloped due to the novelty of the method and its specificities compared to traditional CT. With Carestream Dental as industrial partner, the first part of this work is a new semi-automatic segmentation protocol for teeth, based on shape and intensity constraints, through a graph-cut optimization of an energy formulation. Results show a good quality of segmentation with an average Dice coefficient of 0.958. A fully functional implementation of the algorithm has led to a software available for dentists, taking into account the clinical context leading to temporal and technical difficulties. A preliminary extension to multi-objects segmentation showed the necessity to get more stringent shape constraints as well as a better optimization algorithm to get acceptable computation times. The second part of this thesis, more prospective, is about the creation of a structural model of the maxillo-facial space, to formalize the a priori knowledge on organs and theirs spatial relations. This model is a conceptual graph where structures and relationships are seen as concepts. In particular, the spatial relations “Along” and “Aligned”, modeled in a fuzzy set framework, have been extended to 3D objects.
Complete list of metadatas

Cited literature [139 references]  Display  Hide  Download

https://pastel.archives-ouvertes.fr/tel-02117909
Contributor : Abes Star <>
Submitted on : Thursday, May 2, 2019 - 5:12:52 PM
Last modification on : Wednesday, July 3, 2019 - 3:02:03 PM

File

EvainManuscritFinal.pdf
Version validated by the jury (STAR)

Identifiers

  • HAL Id : tel-02117909, version 1

Citation

Timothée Evain. Nouvelles méthodes de segmentation en imagerie tomographique volumique à faisceau conique dentaire. Imagerie médicale. Télécom ParisTech, 2017. Français. ⟨NNT : 2017ENST0066⟩. ⟨tel-02117909⟩

Share

Metrics

Record views

134

Files downloads

31