Abstract Bruno Torresani, Aix-Marseille Université:
Space-time neuro-electromagnetic inverse problem: trying to take time seriously into account
Neuro-electromagnetic brain imaging aims at inferring the electrical brain activity from external electromagnetic measurements (EEG or MEG). The goal is often reduced to identify "active regions" on the cortical surface (this is the so-called source localization problem), time dependence is often disregarded. However, there is growing interest in time dynamics, and in the inference of time-dependent brain activity networks, and therefore in accurate resolution of the space-time EEG and MEG inverse problems.
In this talk we will describe MEG/EEG inverse problems and the classical approaches, with a focus on space-time problems. We will then discuss our contribution, that exploits simultaneously wavelet modeling of time dependence, probabilistic modeling of space-time data and factorization methods, in the framework of the so-called Maximum Entropy on the Mean paradigm. This will be illustrated by results on sleep MEG data.
This is joint work with MC Roubaud, JM Lina, J. Carrier, M. Lannes and C. Verrier