HAL will be down for maintenance from Friday, June 10 at 4pm through Monday, June 13 at 9am. More information
Skip to Main content Skip to Navigation

Estimation et classification des temps de relaxation multi-exponentiels en IRM. Application aux tissus végétaux

Abstract : Acquired relaxation data in magnetic resonance imaging makes it possible to conduct fine analysis of tissues composition. Conventionally, the analysis is realized by adopting a mono-exponential model at each voxel of the image, yet, a multi-exponential decay model may provide richer information. However, obtaining and interpreting multi-exponential relaxation time maps at the whole image level, from magnitude MRI images, requires solving a large scale inverse problem. This thesis work proposes algorithms for multiexponential relaxation times and their associated intensities maps reconstruction. These algorithms are based on the maximum-likelihood estimator under the hypothesis of a Rician noise distribution, case of magnitude images, and a spatial regularization favoring the regularity of the maps. The resulting large-scale optimization problem is solved using an iterative descent approach by majorization-minimization coupled with a Levenberg-Marquardt algorithm with step search. Finally, we propose a method for image composition characterization from the estimated parameters using classification algorithms. The developed algorithms in this thesis are applied to vegetal tissue analysis.
Complete list of metadata

Cited literature [75 references]  Display  Hide  Download

Contributor : Abes Star :  Contact
Submitted on : Wednesday, September 30, 2020 - 4:00:17 PM
Last modification on : Wednesday, April 27, 2022 - 3:51:11 AM


Version validated by the jury (STAR)


  • HAL Id : tel-02561480, version 2


Christian El Hajj. Estimation et classification des temps de relaxation multi-exponentiels en IRM. Application aux tissus végétaux. Traitement du signal et de l'image [eess.SP]. École centrale de Nantes, 2019. Français. ⟨NNT : 2019ECDN0066⟩. ⟨tel-02561480v2⟩



Record views


Files downloads