Skip to Main content Skip to Navigation
Theses

Modèles de la variabilité géométrique du rachis scoliotique

Abstract : The spine is composed of bony structures (the vertebrae) which are surrounded by soft tissues. Thus, it is a flexible structure, but it is also piecewise rigid. An articulated model enables us to take into account the specific nature of the spine deformations in order to create statistical models of its shape. However, articulated models naturally belong to a Riemannian manifold and not to a vector space. Therefore, the Fréchet mean and a generalized measure of the covariance were used to create statistical models of the spine shape. These statistical models can be intuitively analyzed using specialized visualization tools and compared using hypothesis tests. The analysis of a large group of scoliotic patients revealed that the spine shape variability is inhomogeneous and anisotropic, and that it is possible to localize the vertebral levels that are significantly altered by an orthopedic treatment. A principal component analysis applied in the tangent plane of the Fréchet mean was used to extract the principal deformation modes, which simplifies the analysis of large databases. These principal deformation modes uncovered clinically relevant patterns when applied to scoliotic patients. Finally, a maximum a posteriori estimation procedure can be used to reconstruct 3D spine models from partial radiological data or from 2D landmarks identified on radiographs. Our results indicate that the proposed method outperforms the most commonly used 3D reconstruction method when 2D landmarks are used. Furthermore, a reconstruction error of less than 2mm was obtained when at least 25% of the data were available. The results of our studies indicate that a statistical model of the whole spine geometry, based on an articulated description of the spine, leads to clinically interpretable descriptive statistics and to algorithms that can reconstruct 3D spine models in circumstances where this was previously impossible.
Complete list of metadatas

Cited literature [210 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00630221
Contributor : Service Ist Inria Sophia Antipolis-Méditerranée / I3s <>
Submitted on : Friday, October 7, 2011 - 4:35:06 PM
Last modification on : Monday, October 19, 2020 - 10:55:17 AM
Long-term archiving on: : Sunday, January 8, 2012 - 2:31:28 AM

Identifiers

  • HAL Id : tel-00630221, version 1

Collections

Citation

Jonathan Boisvert. Modèles de la variabilité géométrique du rachis scoliotique. Automatique / Robotique. Université Nice Sophia Antipolis, 2008. Français. ⟨tel-00630221⟩

Share

Metrics

Record views

388

Files downloads

868