Detection and localization of deep sources in magnetoencephalography: a review

Abstract Magnetoencephalography (MEG) is a non-invasive technique for exploring the spatio-temporal dynamics of brain networks with high temporal resolution as well as good spatial capacities thanks to relative insensitivity to low skull conductivity. It is a tool of choice both for neuroscience research and for clinical applications, and is used routinely in epilepsy for localizing the sources of epileptiform discharges. Still, the capacity of capturing deep sources, such as hippocampus and amygdala that are key players in memory and emotion, has been for long a topic of debate. Thus, the fine characterization of deep structures has been up to now the reserved domain of intracerebral EEG, performed during presurgical evaluation of patients with epilepsy. We review here the evidence for the detection of deep sources in MEG, with emphasis on simultaneous recordings of MEG and intracerebral EEG. In this review, we discuss how simultaneously recording depth and surface signals enables to investigate the correlation between MEG and invasive signals actually recorded in deep structures. We also discuss new venues in analysis and recording methods for reconstructing deep activities from MEG.