Periodic EEG features in light sleep and propofol sedation

Frederic von Wegner, Milena Wiemers, Gesine Hermann, Inken Tödt, Enzo Tagliazucchi, Helmut Laufs

DFG project: Objective EEG bed side assessment of impaired consciousness in epilepsy

DGKN 2023

EEG basics

EEG background


  • EEG records electrical activity from several cm2 of cortex
  • each electrode records PSPs of (tens of) millions of neurons
  • volume conduction averages (`blurs`) source activity
  • EEG reflects excitability, neuronal synchrony, and connectivity

Motivation


Example-1

Sleep stage N1

Example-2

Propofol

Preliminary observations

  • EEG appears 'disorganzed' in both drowsy conditions

  • Topographical organization seems to be lost

  • Spectral organization seems to be lost

  • Aim: quantify these observations

Data sets


  • Wake/sleep data set
  • n=32 healthy subjects
  • Simultaneous EEG-fMRI recordings (fMRI not used here)
  • 30 channel EEG (subset of 10-10 system)
  • Sleep stages manually scored (AASM rules)
  • Conditions: wake (W), light sleep (N1)
  • Example publications: Brodbeck et al., 2012, Kuhn et al., 2014 Tagliazucchi et al., 2012, von Wegner et al., 2017

  • Propofol data set
  • n=20 neurologically healthy subjects
  • 128-channel EEG ('clean' subset of 91 channels used)
  • Sedation levels monitored by propofol plasma concentration
  • Conditions: baseline, mild and moderate sedation, recovery. Note: Moderate sedation: still responsive, reaction time delays
  • Published by: Chennu et al., PLoS Comp Biol 2016, data repository: https://www.repository.cam.ac.uk/handle/1810/252736

Main questions

  1. Is EEG activity periodic at the global (scalp-wide) level?

  2. Is periodicity a marker of reduced vigilance (consciousness) ?

EEG microstates (Lehmann, 1972)

How to summarize spatio-temporal information?

Step 1: Microstate maps




Step 2: Microstate sequence


EEG microstates and brain networks



  • Resting-state microstate maps can be mapped to their likely (brain-wide) current sources (Custo et al., Brain Connectivity, 2017)

  • Map C seems to correspond to the default mode network (DMN), others are less clear

  • Take-home: when you see a microstate label (A..D), think "brain network"

Microstates and data compression


  • Initial data set: 30 channels x 16 bit (for instance)

  • Microstate sequence: single letter (A...D) = 2 bit

  • Compression factor: 240

  • Question: can this (0.4%) encode enough information to capture subtle differences in alertness?

Analysis-1 (von Wegner, 2017)

Analysis-2 (von Wegner, in prep.)

Comment

Wiener-Khintchine theorem

Previous results


  • EEG topographies at rest are periodic (von Wegner, NeuroImage 2017, 2022)
  • Their frequency is approx. twice the alpha frequency (20 Hz)
  • Periodicity is quantified by information-theoretical measures
  • Remaining question: frequency doubling? (10 Hz EEG → 20 Hz microstates)

Frequency doubling


  • Explained by the microstate algorithm - ignores polarity
  • EEG polarity inverts every 50 ms (half an alpha phase), both are mapped to microstate C (e.g. 48, 96 ms)
  • Remember: the microstate algorithm doubles the dominant EEG frequency

Results

Question: Loss of consciousness - loss of topography?

Results

Microstate maps in reduced consciousness

Surprise: Microstate maps don't change much in sleep and propofol sedation.

Results


  • Autoinformation analysis - propofol
  • A, baseline (no propofol): microstates occur periodically (50 ms, 100 ms, ...) ~ 20 Hz
  • B, mild sedation (approx. 0.6 µg/ml): periodicity not affected (subjects behaviourally relatively normal, see Chennu et al.)
  • C, moderate sedation (approx. 1.2 µg/ml): periodicity attenuated
  • D, recovery: effects are reversible

Results


  • Spectral analysis - propofol
  • A, baseline (no propofol): EEG main frequency (blue) 10 Hz, microstate sequences (black) 20 Hz
  • B, mild sedation (approx. 0.6 µg/ml): periodicity not affected
  • C, moderate sedation (approx. 1.2 µg/ml): EEG channels have additional beta and delta frequencies, alpha attenuated, microstate frequency peak at 20 Hz disappears
  • D, recovery: effects are reversible

Results


  • Autoinformation analysis - sleep
  • W, wakeful: EEG main frequency (left) 10 Hz, microstate sequences (black) occur periodically (50 ms, 100 ms, ...) ~ 20 Hz
  • N1, light sleep: EEG channels (left) lose their periodicity, microstate sequences (right) too

Results


  • Spectral analysis - sleep
  • W, wakeful: EEG main frequency (left) 10 Hz, microstate sequences ~ 20 Hz
  • N1, light sleep: EEG channels (left) lose 10 Hz alpha periodicity, microstate sequences (right) too

Results

Conclusions

  • EEG microstate sequences capture spatial (topographic) and temporal characteristics simultaneously
  • EEG patterns occur periodically, closely linked to the individual's alpha frequency during wakefulness (explain frequency doubling?)
  • Loss of periodicity as a common feature of reduced consciousness in light sleep and propofol sedation
  • Single electrode level is not a good predictor of global synchrony (propofol, moderate sedation)
  • In the analyzed conditions, periodic, coherent scalp-wide EEG patterns are a signature of intact vigilance (awareness, consciousness)

Explanations & Outlook

  • EEG - biophysical models - coupled oscillators
  • Coupled oscillators close to criticality produce coherent but changing patterns, leading to 'global patterns'
  • Microstates represent momentarily phase-coupled oscillator populations
  • Loss of consciousness - loss of critical coupling
  • Role of information processing and sharing across brain regions

Thank you!

This project is a collaborative effort funded by the DFG

  • Helmut Laufs
  • Gesine Hermann
  • Inken Toedt
  • Enzo Tagliazucchi
  • Frederic von Wegner (f.vonwegner@unsw.edu.au)