Recently, several new beamformers have been introduced for reconstruction and localization of neural sources from EEG and MEG. Although studies have compared the accuracy of beamformers for localization of strong sources in the brain, a comparison of new and conventional beamformers for time-course reconstruction of a desired source has not been previously undertaken. In this study, 8 beamformers were examined with respect to several parameters, including variations in depth, orientation, magnitude, and frequency of the simulated source to determine their (i) effectiveness at time-course reconstruction of the sources, and (ii) stability of their performances with respect to the input changes. The spatial and directional pass-bands of the beamformers were estimated via simulated and real EEG sources to determine spatial resolution. White-noise spatial maps of the beamformers were calculated to show which beamformers have a location bias. Simulated EEG data were produced by projection via forward head modelling of simulated sources onto scalp electrodes, then superimposed on real background EEG. Real EEG was recorded from a patient with essential tremor and deep brain implanted electrodes. Gain - the ratio of SNR of the reconstructed time-course to the input SNR - was the primary measure of performance of the beamformers. Overall, minimum-variance beamformers had higher Gains and superior spatial resolution to those of the minimum-norm beamformers, although their performance was more sensitive to changes in magnitude, depth, and frequency of the simulated source. White-noise spatial maps showed that several, but not all, beamformers have an undesirable location bias.
- Region of interest (ROI)
- Signal-to-noise power ratio (SNR)
- Time-course reconstruction