Figure 1.
Experiment and analysis outline.
(A) EEG was recorded with the extended 10–20 montage, then subsampled to 42 electrodes to directly compare common with Laplacian referenced data. Full, frontal and parietal networks were derived from all, red, and blue channels, respectively. (B) In addition to a 5 minutes eyes closed resting condition, participants underwent N2O gas inspiration over a 20 minute period with peak gas levels of 20, 40 or 60% ( is the number of participants in each group). (C) EEG was recorded with a common linked mastoids reference sensitive to all spatial frequencies as seen in the potential map (left). EEG was also analysed with Laplacian re-referencing which is sensitive to high spatial frequencies (right). The potential maps correspond to snap shots of resting data and are presented just to illustrate the different spatial scales of the signals. (D) Global Efficiency (GE) and Global Coherence (GC) analyses were applied to common (left) and Laplacian (right) referenced data for full, frontal and parietal brain networks to assess global changes in functional connectivity resulting from N2O induction. Examples are given for a participant in the 60% peak gas group for the full brain network. For GE, network graphs are constructed based on surrogate-corrected zero-lag correlations. The graphs correspond to times of initial and peak gas inspiration. The thickness of the graph lines are proportional to the strength of the absolute value of the correlations which vary between 0 and 1. Only correlations of absolute magnitude greater than the 90th percentile value are shown. Red, blue and gray graph edges correspond to electrode pairs involving frontal, parietal, and neither frontal nor parietal electrodes, respectively. Note the decreases in parietal correlations with increased gas concentration for the Laplacian referenced derivation (right), consistent with the hypothesis of a ‘final common pathway’ to drug-induced reductions in consciousness. GE (green time series) for these weighted networks is the average of the absolute correlations. GC is the ratio of the largest eigenvalue over the sum of the eigenvalues of the complex cross-spectral matrix at each temporal frequency, this is reflected in the GC spectra (bottom). Vertical white spaces indicate time intervals in which data contained artefact. Additional measures were also derived from the GE and GC analyses to assess changes in functional connectivity.
Figure 2.
Time series of the GE-based functional connectivity measures for the four different subjects (A–D) from the 60% peak gas group for the parietal networks obtained with Laplacian re-referencing.
In each sub-figure: GE (; green), the contribution of connection strength to GE (
; blue), and the contribution of network topology to GE (
; red). Black curve: measured end-tidal N2O gas concentrations. Magenta curve: smoothed auditory task performance. Missing data indicates artefact.
Figure 3.
Time series of the GE-based functional connectivity measures for the same subject in Figure 1D from the 60% peak gas group for the full brain (top row), frontal (middle row) and parietal (bottom row) networks obtained either with common-reference (left column) or Laplacian re-referencing (right column).
In each sub-figure: GE (; green), the contribution of connection strength to GE (
; blue), and the contribution of network topology to GE (
; red). Black curve: measured end-tidal N2O gas concentrations. Magenta curve: smoothed auditory task performance. The bottom right sub-figure corresponds to Figure 2A. Missing data indicates artefact.
Figure 4.
GC-based functional connectivity spectra for the same subject in Figure 3 for the full brain (top row), frontal (middle row) and parietal (bottom row) networks obtained either with common-reference (left column) or Laplacian re-referencing (right column).
In each sub-figure: Black curve - measured end-tidal N2O gas concentrations; brown curve - smoothed auditory task performance. Vertical white spaces indicate time intervals in which data contained artefact.
Figure 5.
Dependence, for the 60% peak gas group, of GE-based functional connectivity () defined relative to the median during rest on N2O gas concentration for full brain (top row), frontal (middle row) and parietal (bottom row) networks, obtained either with common-reference (left column) or Laplacian re-referencing (right column).
Box-whisker details: Red horizontal lines indicate bin distribution median; Upper and lower blue box edges correspond to 75th and 25th percentiles, respectively; Black dashed whiskers span 99.3% of the distribution assuming a normal distribution; The red crosses indicate outliers. The bins labelled above by and
indicate the mean is statistically significant from rest for significance levels of
and
corrected for multiple comparisons, respectively.
Figure 6.
Dependence, for the 60% peak gas group, of GC-based functional connectivity () at 11 Hz defined relative to the median during rest on N2O gas concentration for full brain (top row), frontal (middle row) and parietal (bottom row) networks, obtained either with common-reference (left column) or Laplacian re-referencing (right column).
Box-whisker and multi-comparison test significance marker (,
) details are the same as for Figure 5.
Figure 7.
AUROC as a function of inspired N2O concentration for the 60% peak gas group for (top left) weighted global efficiency, , (top right) global coherence,
, at 11 Hz, (bottom left) the contribution of connection strength to global efficiency,
, and (bottom right) the contribution of connection topology to global efficiency,
.
One-way ANOVAs were typically significant across rest and the six gas concentration bins for each of the measures ( for the GE measures,
for the GC measure except for parietal networks with Laplacian-reference derivation). The difference in the median measure value relative to rest is indicated by the the direction of the respective bar (up, increase; down, decrease). Gray shading indicates ROC curve areas
signifying a chance or worse ability to discriminate between randomly chosen gas and rest measures. Thus AUROC
for a given measure implies a better than chance likelihood of being able to discriminate between a gas state and rest. Multi-comparison test significance marker (
,
) details are the same as for Figure 5.
Figure 8.
AUROC as a function of aCPT accuracy for the 60% peak gas group for (top left) weighted global efficiency, , (top right) global coherence,
, at 11 Hz, (bottom left) the contribution of connection strength to global efficiency,
, and (bottom right) the contribution of connection topology to global efficiency,
.
One-way ANOVAs were typically significant across rest and the two accuracy bins for each of the measures ( for the GE measures except for parietal
and
with common-reference derivation,
for the GC measure except for the parietal network with Laplacian-reference derivation). The difference in the median measure value relative to rest is indicated by the the direction of the respective bar (up, increase; down, decrease). Multi-comparison test significance marker (
,
) details are the same as for Figure 5. The remaining features are the same as described in Figure 7.
Figure 9.
AUROC as a function of aCPT reaction time for the 60% peak gas group for (top left) weighted global efficiency, , (top right) global coherence,
, at 11 Hz, (bottom left) the contribution of connection strength to global efficiency,
, and (bottom right) the contribution of connection topology to global efficiency,
.
One-way ANOVAs were typically significant across rest and the five reaction time bins for each of the measures ( for the GE measures except for frontal
with common-reference derivation,
for the GC measure). The difference in the median measure value relative to rest is indicated by the the direction of the respective bar (up, increase; down, decrease). Multi-comparison test significance marker (
,
) details are the same as for Figure 5. The remaining features are the same as described in Figure 7.