Study of experimental conditions relevance in a Bayesian framework for functional MRI analysis: PhD (3 years, 2010-2013)

The context of our study is the brain activity detection and HRF estimation on the whole brain. To this end, Makni et al. 2005 and Makni et al. 2008 have introduced a joint detection estimation (JDE) framework based on a Bayesian General Linear Model (GLM) that implements such a joint procedure on a prior parcellation of the function brain mask. This mean that the JDE methodology consists in alternating HRF estimation with activity detection on each parcel separatly. Vincent et al. 2009 and Risser et al. 2009 have considered spatially adaptive mixture models to incorporate spatial correlation within each parcel but not between adjacent ones. Inference is based on simulation intensive Markov Chain Monte Carlo (MCMC) methods. The most recent extensions of the JDE framework (Lotfi et al. 2010) showed the advantages of the variational approaches (VEM) versus MCMC, especially the gain in computation time. Studies usually assume that all delivered stimuli possibly generate a BOLD response everywhere in the brain although activation is likely to be induced by only some of them in specific brain areas. The goal of this work is to implement a variable selection procedure that automatically selects the conditions according to the brain activity they elicit.

Estimation of respiratory movements using the elastic Optical Flow registration: M. Sc. thesis (6 months in 2010)

Respiratory motion affects the qualitative and quantitative accuracy of multimodality PET/CT (Positron emission tomography/Computed tomography) images as well as the dose delivered to tumor and healthy tissues during radiation therapy (RT). The optical flow is an elastic registration algorithm allowing the computation of the deformation vectors in a 4D CT image sequence. This allows us to link the tumor and internal organs motion to external markers and thus improve the planning and dose delivery during radiotherapy.