Spatial spectra of scalp EEG and EMG from awake humans

Abstract

Objective: Evaluate spectral scaling properties of scalp electroencephalogram (EEG) and electromyogram (EMG), optimal spacing of electrodes, and strategies for mitigating EMG.

Methods: EEG was recorded referentially from 9 subjects with a 64 channel linear array (electrodes 3mm apart) placed parasagittally or transversely on forehead or occiput, at rest with eyes open or closed, or with deliberate EMG. Temporal (PSD(t)) and spatial (PSD(x)) power spectral densities were calculated with one-dimensional fast Fourier transform (FFT) for comparison with earlier analyses of intracranial EEG.

Results: Scaling of PSD(t) from scalp resembled that from pia: near-linear decrease in log power with increasing log frequency (1/f(alpha)). Scalp PSD(x) decreased non-linearly and more rapidly than PSD(x) from pia. Peaks in PSD(t) (especially 4-12Hz) and PSD(x) (especially 0.1-0.4 cycles/cm) revealed departures from 1/f(alpha). EMG power in PSD(t) was more "white" than 1/f(alpha).

Conclusions: Smearing by dura-skull-scalp distorts PSD(x) more than PSD(t) of scalp EEG from 1/f(alpha) scaling at the pia. Spatial spectral peaks suggest that optimal scalp electrode spacing might be approximately 1cm to capture non-local EEG components having the texture of gyri. Mitigation of EMG by filtering is unsatisfactory. A criterion for measuring EMG may support biofeedback for training subjects to reduce their EMG.

Significance: High-density recording and log-log spectral display of EEG provide a foundation for holist studies of global human brain function, as an alternative to network approaches that decompose EEG into localized, modular signals for correlation and coherence.