A systematic data-driven approach to analyze sensor-level EEG connectivity: Identifying robust phase-synchronized...

Area of Science:

• Neuroscience
• Brain Network Analysis
• Electroencephalography (EEG)

Background:

• Conventional electroencephalography (EEG) functional connectivity (FC) analysis simplifies data by averaging across frequency bands, but this obscures resting-state brain network (RSN) identification and hinders FC reliability estimation.
• Existing methods struggle to accurately capture the dynamic and reliable nature of brain networks derived from EEG.
• There is a need for advanced analytical techniques to improve the precision and interpretability of EEG-based FC.

Purpose of the Study:

• To develop and validate a novel approach combining scalp current source density (CSD) transformation and spectral-spatial principal component analysis (PCA) for identifying robust and reliable functional connectivity components in resting-state EEG.
• To demonstrate that this method can identify specific resting-state brain networks (RSNs) with high accuracy and reliability.
• To assess the sensitivity of identified FC components to experimental conditions like eyes open/closed.

Main Methods:

• Resting-state EEG data from 35 healthy adults (71 sensors, 8 minutes, eyes open/closed, 1-week retest) were analyzed.
• EEG data were transformed using scalp current source density (CSD) and phase-based FC was estimated using a debiased-weighted phase-locking index.
• Spectral-spatial PCA was employed to extract and identify robust FC components, followed by internal consistency and test-retest reliability assessments.

Main Results:

• Spectral PCA identified six robust alpha and theta components capturing 86.6% of the variance.
• Subsequent spatial PCA revealed seven spatially distinct alpha and theta FC components, consistent with known RSNs (e.g., default mode, visual, sensorimotor), accounting for 37.0% of FC variance.
• Four identified components were sensitive to eyes open/closed conditions, and most components demonstrated good-to-excellent internal consistency and test-retest reliability (ICCs ≥ 0.8).

Conclusions:

• The proposed CSD and spectral-spatial PCA method effectively reduces EEG functional connectivity dimensionality, yielding meaningful and reliable network components without arbitrary thresholds.
• Identified FC components align with established RSNs and exhibit high reliability, suggesting their utility for basic and clinical neuroscience research.
• This approach offers a significant advancement for accurately quantifying brain network dynamics from EEG data.