Resume:

I have an Engineering degree from ESME Sudria (1999) with a major in electronics and signal processing and a Master's degree from Institut National Polytechnique de Grenoble (INPG, 2000) in signal and image processing.

I defended my PhD in september 2003 at the University of Paris V. The topic of my thesis was "Non rigid registration using statistical variational approaches. Application to the analysis and the modelling of the myocardial function in MRI", a work carried out under the supervision of Prof. Françoise Prêteux and Nicolas Rougon at ARTEMIS, at Telecom SudParis (ex-Institut National des Télécommunications, INT).

Before joining University of Rouen in September 2005, I worked in a Contract Research Organization (CRO) as a research engineer (2003-04) and held a post-doctoral position (2005) at the Centre des Mathématiques et de Leurs Applications (CMLA) at ENS Cachan.

Summary of my PhD thesis (defended in 2003):

Quantitatively assessing the myocardial contractile function is a major issue for the prevention, treatment and follow-up of cardiovascular diseases. In this context, Magnetic Resonance Imaging (MRI) is a privileged modality for dynamically exploring the heart. However, the quantitative exploitation of MRI data remains limited today, due to the lack of reliable, robust and reproducible non rigid motion estimation techniques from MR image sequences.
This thesis aims at demonstrating that statistical non rigid registration techniques define an appropriate framework for quantitatively assessing myocardial deformations. Its contributions concern:

• the elaboration of a robust unsupervised method for estimating myocardial motion from tagged MR sequences. It allows to derive reliable pointwise displacement measurements within the myocardium at every cardiac phase and slice level.
• the development of a measurement tool for dynamically quantifying myocardial deformations, integrating an automated heart segmentation procedure via non rigid registration of cine MR data. For the healthy heart, comparing the resulting measurements with reference values derived from an in-depth synthesis of medical literature demonstrates an excellent correlation. For pathological hearts, experiments have illustrated the relevance of a multi-parametric quantitative analysis for localizing and characterizing diseased areas.
• the construction of a statistical atlas for the healthy heart, which provides a numerical reference for normality and can be used as a parametric motion model to drive the non rigid registration process of arbitrary tagged MR data. A very compact description of myocardial displacements is then obtained without notable accuracy loss.

pdf download (in french)