Improved beamformer with weighted source region suppression for coherent MEG source localization

摘要

Beamformer is one of the main techniques for spatio-temporal neuroelectromagnetic source reconstruction. However, the classical Beamformer is extremely sensitive to strongly coherent sources, thereby encountering difficulty in localizing the highly correlated bilateral auditory cortices in auditory evoked field (AEF) or auditory steady state evoked potential. The multiple constrained minimum-variance Beamformer with coherent source region suppression (Beamformer-CS) can potentially overcome such difficulties. However, when coherent interferer is located close to the edges of the suppression region, Beamformer-CS has localization bias and the closer it is, the larger it will be. Here, we present an improved Beamformer-CS that can localize coherent sources with much less localization bias, especially in the case of the interferer close to the edges of the suppression region. First, based on approximate information about source energy distribution from other neuroimaging techniques, a region encompassing the coherent interfering sources is defined. Then, the dominant eigenvectors of the lead field matrix, weighted using source energy information obtained by other imaging method, for the suppression region is incorporated into Beamformer design as hard null constraints. Such weighting strategy is able to improve the localization performance. Simulation test shows that, compared to Beamformer-CS, the new weighting approach is of much smaller localization bias, sharper peak of the estimated sources, more robust against noise, and less sensitiveness to the number of the eigenvector components for the suppression region, as is also confirmed by real AEF data test.