Reproducibility of power spectrum, functional connectivity and network construction in resting-state EEG

Characteristics from resting-state electroencephalography (rsEEG) provides relevant information about individual differences in cognitive tasks and personality traits. Due to its increasing application, it is crucial to know the reproducibility of several analysis measures of rsEEG.

A new brain network construction method was proposed based on simplified forward model (SFM). In addition, we aimed to carry out an extensive examination of the reproducibility of the power spectrum and functional connectivity at both the sensor-level and the source-level. We systematically proposed multiple new pipelines by integration source imaging, time-course extraction and network reconstruction.

Our results revealed that the reproducibility of eyes-closed was slightly higher than that of eyes-open, and the relative power was more repeatable than the absolute power, especially in high-frequency bands. The reproducibility of the sensor-level was higher than that of the source-level, both for power and connectivity. Remarkably, connectivity measures could be separated into two classes according to their reproducibility. Notably, the reproducibility of power envelope correlation (PEC) was generally the highest among those connectivity measures which are insensitive to volume conduction effect. For the whole-brain network construction, single dipole modeling was better than the dimensionality reduction methods, such as mean or principal component analysis (PCA) of multiple dipoles of a region.

In conclusion, our results described the reproducibility of rsEEG power spectrum, connectivity measures, and network constructions, which could be considered in assessing inter-individual differences in brain-behavior relationships, as well as automatic biometric applications.