Characteristic changes in the EEG signals between microsleeps and preceding responsive states

Area of Science:

• Neuroscience
• Signal Processing
• Computational Biology

Background:

• Microsleeps are brief, involuntary disruptions of consciousness.
• Distinguishing microsleeps from attentive states using EEG is crucial for safety-critical applications.
• Current EEG analysis methods may not fully capture the dynamic network changes during microsleeps.

Purpose of the Study:

• To identify characteristic EEG features for differentiating microsleeps from preceding responsive states.
• To investigate changes in brain network organization during microsleeps using graph theory.
• To explore potential EEG-based biomarkers for microsleep detection.

Main Methods:

• EEG signals were re-referenced using the reference electrode standardization technique (REST).
• Time-varying auto-regressive (TVAR) parameters were derived using a time-varying general linear Kalman filter (TVGLKF).
• Time-varying effective connectivity was quantified using orthogonal partial directed coherence (OPDC) and analyzed with graph theory, including community-based measures.

Main Results:

• A significant decrease in directional modularity from anterior to posterior was observed in theta, alpha, and beta bands during microsleeps.
• The alpha band exhibited the highest significance (Cohen-type effect size = 1.25), with a median percentage difference of 23%.
• Decreased flexibility and integration, alongside increased recruitment, were observed in brain networks during microsleeps across significant bands.

Conclusions:

• Community-based graph theory measures derived from EEG effective connectivity can characterize brain mechanism changes during microsleeps.
• These measures show potential as reliable biomarkers for objective microsleep detection.
• The findings provide insights into the neural dynamics underlying microsleep transitions.