Super-resolution of multi-contrast MRI using variational

Ajouter à mon calendrier
Google Agenda Outlook Calendrier

Abstract  

Obtaining high-resolution isotropic MRI images required for the diagnosis of certain conditions (joint disorders, neuroimaging, colorectal cancer) can result in long acquisition times. Generally, a trade-off must be made between acquisition time, signal-to-noise ratio, and spatial resolution. In this thesis, we address the technique of MRI super-resolution, which allows for the generation of high-resolution isotropic volumes from low-resolution anisotropic acquisitions while controlling acquisition time and preserving the signal-to-noise ratio. In this thesis, research was conducted from a variational perspective, considering two cases: single-contrast super-resolution and multi-contrast super-resolution. Several in vivo acquisition protocols were conducted on healthy volunteers. Throughout the project, the use of the SPACE sequence developed by Siemens Healthcare played a very important role not only in performing the various experiments but also in conducting quantitative and qualitative evaluations during image reconstruction. In the single-contrast case, the main contribution was the application of a discrete Mumford-Shah model in the context of MRI super-resolution. In the multi-contrast section, the work consisted of extending the super-resolution methods initially designed for the single-contrast case to account for anisotropic acquisitions with different contrasts. The main contribution of the thesis was to provide a multi-contrast super-resolution method that allows for the combination of anisotropic volumes with different contrasts and orientations to reconstruct a high-resolution isotropic volume by contrast without resorting to an isotropic reference volume.

KEYWORDS : Super-resolution, multi-contrast MRI, variational methods.

 

Jury

Freddy Odille ,  DR, Inserm, Rapporteur. 

Denis Kouamé, PU, Université Paul Sabatier, Rapporteur.                      

Olivier Beuf, DR, CNRS, Examinateur. 

François Rousseau, PU, IMT, Examinateur.

Marylène Delcey, IR,  Siemens Healthcare SAS, invitée.

Hélène Ratiney, DR, CNRS, Directrice de thèse.

Eric Van Reeth,  Enseignant-chercheur, CPE, Co-encadrant de thèse.

Marion Foare, Enseignant-chercheur, CPE, Co-encadrante de thèse, invitée.