Adaptive estimation strategy for coupled EEG-MEG analysis

Human brain electric activity related to important physiological functions can be pictured from bio electric and/or bio magnetic measurements through inverse problems approach. In this paper an iterative strategy is presented. The strategy is optimized to reconstruct brain activity confined in a sub region of the brain volume by jointly processing both bioelectric and biomagnetic signals. The procedure takes advantage from intermediate evaluations to improve the resolution, while limiting the number of unknowns. The advantages of the proposed approach is a higher robustness against noise and uncertainty and better effectiveness. In the last few decades, important contributions in the neuroscience research came from the use of imaging techniques. In particular, new investigation approaches on the anatomy of the brain, based on structuralimagingtechniques,suchascomputeraidedtomography(CT)andmagneticresonanceimaging (MRI), have been proposed. A major field of research in neuroscience is aimed to understand metabolic processes taking place in the brain; this calls for several imaging techniques such as positron emission tomography (PET), single-photon emission tomography (SPECT) and functional magnetic resonance imaging (fMRI) measuring changes of the blood flow or oxygenation in the brain. Such techniques are abletomeasurebrainactivitywithsatisfactoryspatialresolution;unfortunatelythetimeresolutionisquite poor becauseit is limited within the range 0.1 s-1 s. Therefore further techniquescharacterizedby better time resolution, are looked for in the scientific community gathering contributions from mathematical, physicalandengineeringdisciplines. Thesetechniques(1-6)includeelectroencephalography(EEG)and magnetoencephalography(MEG). Electrical currents in the brain can be identified by processingvoltage patterns measured on the scalp (EEG) and/or by measuring the extremely weak magnetic flux density close to the scalp (MEG). The MEG has been recently introduced thanks to the development of the SQUID (Superconducting Quantum Interference Devices) sensors, able to detect the very low magnetic flux density in the extracranial region, in the order of a few fT. Compared with fMRI techniques, EEG and MEG represent an appealing alternative thanks to the time resolution (in the order of 1 ms), and to the lack of radiations (1). Unfortunately their spatial resolution could be not satisfactory (1). In addition, due to the ill posedness of the model, the uncertainty affecting the knowledge of some of the electrical