This research is focused on the application of statistical methods (eg, Bayesian inference) to. The Joint Detection-Estimation (JDE) framework I have promoted gives access the neuroscientist with region-specific HRFs and detection maps of brain activity in response to given experimental task or stimulus or to a combination of those from functional brain imaging (fMRI) data. This novel modelling takes place in the Bayesian formalism and is flexible enough to account for some putative deactivations (eg epilepsy) or neuronal habituation generating the so-called repetition-suppression effect. Moreover, spatial dependency is accounted for through the use of Markov random fields that avoids us to spatially filter the data sets.
Currently, we are working on the following directions:
- The impact of the JDE framework on group-level studies in fMRI by examining a huge datasets composed of different acquisitions for each subject, each acquisition targeting a different spatial in-plane resolution in GE BOLD EPI images. This part is managed by Solveig Badillo and Thomas Vincent. See ISBI'11 papers related to this issue (go to LNAO web site, section Bibliography).
- a Variational Bayes alternative to the sampling-based strategy for computing the Posterior Mean estimates and studying the accuracy of the associated results regarding the involved factorization in the approximation. This study is conducted in close collaboration with Florence Forbes. We jointly supervise Lotfi Châari during his INRIA-funded post-doc position on this topic. This work has been published in the 2011 MICCAI conference [Pubmed reference, pdf see Attachment below].
Our collaborative research project named AINSI
and coding for Modèles statistiques pour l’Assimilation d’Informations de Neuroimagerie fonctionnelle et de perfuSIon cérébrale
been funded for two years. More details are available on our specific web site