Неинвазивное картирование коррелированных источников методом модифицированного формирователя луча

MEG is a non-invasive neuroimaging modality hallmarked by the order of millisecond temporal and sub-centimeter spatial resolution, which makes it an ideal tool to study subtle spatial and temporal brainworkings dynamics. High temporal resolution of MEG comes from the nature of electric processes it registers. The spatial resolution crucially depends on the approach to the ill-posed inverse problem aimed at finding cortical activations based on the non-invasive measurements of the magnetic fields. Linearly constrained minimum variance (LCMV) beamforming technique possess spatial super-resolution and is therefore an attractive tool to deal with the MEG inverse problem. The technique is based on adaptive spatial filter design with goal to minimize output power subject to a set of linear constraints. In the presence of correlated sources this scheme fails and results into spatial filter with the poor performance, low SNR of the output timeseries and meaningless spatial maps of power distribution. With this paper a framework for ReciPSIICOS, a novel beamformer approach, was proposed. This technique preserves spatial super-resolution but remains immune to the synchrony of neuronal source timeseries by emphasizing the power in the spatial leakage subspace. ReciPSIICOS algorithm was tested along with LCMV beamformer and weighted MNE source reconstruction techniques using both realistic simulation and real data. As simulations and real data analysis show, novel beamformer has greater tolerance for high signal correlation and low signal-to-noise ratio.