The Performance Of The MEG Source Localization Based On Beamforming

Beamforming is a kind of method widely used in the dipole sourcing of MEG(Magnetoencephalography) signals. However, the localization effect is affected by various complex factors, so improving the accuracy of localization is the main work in recent study. By MATLAB simulation, this paper studies the factors of two kinds of beamforming methods: dynamic imaging of coherent source(DICS) and linear constrained minimum variance(LCMV), obtains the quantitative conclusions, which have some guiding significance for source localization to real MEG data.First, from the basic points of biological electromagnetic research, this paper explain the methods and modelling for solving the forward and inverse problem in MEG source calculation. Then some common beamforming algorithms are emphasized and compared.Second, the influence of the regularization parameters in DICS is studied by adjusting the regularization values in localization of the simulation MEG data with various artifacts. The result shows that the regularization parameter can suppress the various artifacts with different degrees. The various artifacts can be suppressed while a more stable localization result can be obtained, when the regularization parameter value is set in the range of 0.001% to 1%.Third, the effect of the different artifacts and head models in LCMV source localization and reconstruction is analyzed. A variety of artifacts are added to the simulated MEG signals which are calculated through some head models in the forward problem. Then the LCMV sourcing is simulated in MATLAB. The simulation suggests that the localization result isn’t affected by the head model when the signal to noise ratio reaches a threshold. Furthermore, the types of artifacts accept different results, where baseline wander artifact has the least influence while the Gaussian white noise has the biggest influence. Overall, the LCMV can show a good performance when the SNR is higher than-10 d B.Finally, according to the conclusion of the simulation above, the DICS and the LCMV methods are applied to the real MEG signals for source localization. It shows that the two beamforming algorithms can both solve the inverse problem of real MEG data reliably under the relevant parameter setting related to the simulation.