[Objective] This study was aimed to evaluate our previously proposed method of estimating the target signal sources in the human brain, with using the electroencephalogram (EEG) data during a simple motor task. And the locations and frequencies of the signal sources estimated by our method are compared with results of the conventional event-related de-/synchronization (ERD/ERS) analysis.
[Method] Eight healthy male volunteers were asked to perform a voluntary finger single-tapping every around 10 s, but without counting, for 5 min per a session; two sessions were made with in each of the subjects. EEG signals were recorded from 19 scalp locations according to the International 10-20 EEG System with a linked earlobe reference. The electromyogram (EMG) was also recorded from the right first dorsal interosseous muscle. Both EEG and EMG data were sectioned by a 1-s Hanning window with 50% overlap for the Fourier transform. The Spearman’s rank-order correlation coefficients in the frequency domain were calculated for all spectral pairs between the EEG signal of each sensor and the EMG signal. Hence, the cross-frequency EEG-EMG correlation matrices were obtained for the 19 EEG sensors. For the spectral pairs showing the statistically significant correlation, spatial distributions of the correlation coefficients over the scalp were plotted as the correlation topographies. Also, the corresponding EMG spectra were extracted as the time series. By these EEG-EMG correlation topographies and the EMG spectra time series, the original EEG data were weighted so that the specific frequency signals were exclusively enhanced. On the weighted EEG data, the standardized low-resolution brain electromagnetic tomography (sLORETA) was applied to each of 1-s time sections, and the source locations were estimated by averaging the sources calculated for all sections with sLORETA modified for quantitative method (sLORETA-qm), using a four-layer anisotropic spheres model with a reference MR image. For the ERD/ERS analysis, the event timing of the finger tapping was determined based on the EMG signal. The EEG signals were extracted for a time frame of a 2.5-s pre-event and a 5-s post-event and the short-time Fourier transform was computed for the spectral power time series.
[Results] The cross-frequency EEG-EMG correlation matrices showed that distinct alpha-band and beta-band EEG signals negatively correlated with the EMG signals by a statistically significant amount. Corresponding to those frequency bands, multiple signal sources were estimated by our proposed method: alpha-band signal sources located in parietal-occipital as well as temporal lobes, and beta-band signal sources located in contra-/ipsilateral regions around central sulcus. The ERD/ERS analysis demonstrated also the alpha-band post-event synchronizations in parietal-occipital lobe, the beta-band desynchronization during the muscle contraction in ipsilateral central region, and the beta-band post-event synchronizations in contralateral central-parietal regions.
[Conclusion] It was shown that our proposed method of the targeted signal source estimation can be useful for localizing the specific neural activities in human brains. And the frequencies and locations of the estimated signal sources were consistent with previous anatomical and physiological studies, and also with the results of conventional ERD/ERS analysis.
View full abstract