Associations between spontaneous electroencephalogram oscillations and oxygen saturation across normobaric and hypobaric hypoxia

Figures

• 

  FIGURE 1. All three conditions were recorded within the hypobaric chamber. The chamber can be used for hyperbaric or hypobaric experiments, and for this study, it was in hypobaric mode. The NN and NH conditions were performed at 334 m.

• 

  FIGURE 2. (a) View of participant set up from the side. The 64‐channel EEG cap is worn on the participant's head, with the mask straps going over the electrodes to maintain an adequate seal on the mask. An SpO2 monitor can be seen on the left index finger. A computer was positioned 60 cm in front of the participant's eyes that was used to provide a fixation cross for data collection during resting‐state conditions, and a subsequent cognitive function task. Note that the results of the cognitive function task are beyond the scope of the present study, which examines responses during eyes‐open resting‐state conditions. (b) Experimental set up seen from the back. The chamber tender was behind the participant and manually matched the flow rate of the gas (either hypoxic or ambient mixture) to the participant's breathing rate. The two laptops to the right of the participant were used to record the EEG and SpO2 data.

• 

  FIGURE 3. Oxygen saturation averaged over the three conditions. The three conditions were significantly different from each other, with NN having the highest SpO2 values, followed by NH, and HH had the lowest SpO2 values (p ≤ .0015). Data are presented as mean ± standard deviation. HH, hypobaric hypoxia; NH, normobaric hypoxia; NN, normobaric normoxia.

• 

  FIGURE 4. Participant‐specific overall correlations between EEG measures and SpO2. Each value is associated with one PLS analysis performed for one EEG measure and one participant (13 in total) in one of the three conditions. More specifically, behavioral PLS analysis was performed on each participant (i.e., 1–13) for absolute spectral power, relative spectral power, and MSE correlated with SpO2. The x axis stands for participants, whereas the y axis shows EEG metrics within conditions. The color bar shows the range of participant‐specific overall correlations between 0.29 and 0.81.

• 

  FIGURE 5. Group‐averaged correlations between EEG rhythms and oxygenation in three conditions for three EEG metrics: absolute spectral power, relative spectral power, and MSE. The correlations were averaged across participants, separately for each condition. Three subplots in one column correspond to one condition. Red values constitute a positive correlation (decrease in power with decrease in SpO2) and blue values represent a negative correlation (increase in power with a decrease in SpO2). Across the three measures, NH has stronger correlations with SpO2 than NN and HH when averaged across individuals. (a) Frequency source matrices for absolute power correlations with SpO2. The NH condition exhibits an increase in absolute power across sources with a decrease in SpO2. (b) Frequency source matrices for relative power correlations with SpO2. A prominent increase in relative power at roughly 8–12 Hz is found across sources with a decrease in SpO2 in the NH condition. (c) Frequency source matrices for MSE correlations with SpO2. The NH conditions exhibit an increase and decrease across sources in sample entropy in finer time scales with a decrease in SpO2.

• 

  FIGURE 6. Group‐averaged correlations between EEG rhythms and SpO2 in NH for three EEG metrics: absolute spectral power, relative spectral power, and MSE. These are the same maps of correlations, which are shown in Figure 5 (three subplots in the right column), only this time we show them projected to the cortical surface. (a) Absolute power negatively correlated with SpO2 exhibits an increase in absolute power with a decrease in SpO2, in the alpha frequency band, alongside the low and high beta frequencies. (b) As for the relative spectral power, a negative correlation between alpha power and SpO2 alongside positive correlations between delta and theta power with SpO2 is exhibited. These correlations show a redistribution of power toward the alpha band and away from the delta and theta bands (i.e., alpha relative power increases as SpO2 decreases, and theta and delta relative power decrease as SpO2 decreases). (c) Lastly, MSE exhibits negative and positive correlations with SpO2, showing a dissociation of frontal and poster regions in terms of sample entropy at finer time scales (i.e., time scales 1–6), with an increase in sample entropy shown over the frontal lobe and a decrease in…

• 

  FIGURE 7. Differences in the correlations between EEG absolute power and SpO2 across the three conditions at the group level, as revealed by the mean‐centered PLS analysis. (a) An overall contrast across the conditions and the corresponding map of z‐scores, each associated with a unique combination of a frequency and brain region. The contrast on the left subplot differentiates NH versus NN and HH (p = .001). Large in magnitude negative z‐scores (shown in blue) robustly support the overall contrast, indicating lower (and in this case, more negative) correlations between EEG absolute power and SpO2 in the NH condition. (b) The same map of z‐scores, as in (a), but with the z‐scores shown on the brain template. The alpha rhythm exhibits the greatest effect due to hypoxia, suggesting global inhibition or a global idle brain state.

• 

  FIGURE 8. Differences in the correlations between EEG relative power and SpO2 across the three conditions at the group level, as revealed by the mean‐centered PLS analysis. (a) An overall contrast across the conditions and the corresponding map of z‐scores, each associated with a unique combination of a frequency and brain region. The contrast on the left subplot differentiates NH versus NN and HH (p = .014). Large in magnitude negative z‐scores (shown in blue) robustly support the overall contrast, indicating lower (and in this case, more negative) correlations between EEG relative power and SpO2 in the NH condition. Positive z‐scores (shown in red) indicate higher and more positive correlations between EEG relative power and SpO2. (b) The same map of z‐scores, as in (a), but with the z‐scores shown on the brain template. A shift in power is observed toward the alpha band and away from the slower frequency rhythms.

• 

  FIGURE 9. Differences in the correlations between MSE of EEG signals and SpO2 across the three conditions at the group level, as revealed by the mean‐centered PLS analysis. (A) An overall contrast across the conditions and the corresponding map of z‐scores, each associated with a unique combination of a frequency and brain region. The contrast on the left subplot differentiates NH versus NN and HH (p = .043). Large in magnitude negative z‐scores (shown in blue) robustly support the overall contrast, indicating lower (and in this case, more negative) correlations between MSE and SpO2 in the NH condition. Positive z‐scores (shown in red) indicate higher and more positive correlations between MSE and SpO2. (b) The same map of z‐scores, as in (a), but with the z‐scores shown on the brain template. The MSE results seem to show a more heterogenous spatiotemporal pattern of correlations than absolute and relative power and may be a more sensitive measure to changes related to hypoxia. At finer time scales (scales 2–6), we see an increase in sample entropy in the frontal regions (negative correlation between MSE and SpO2) and decrease in poster regions (positive correlation between MSE and SpO2).…

• 

  FIGURE 10. Differences in the correlations between EEG absolute power and SpO2 across the three conditions at the group level split into the first and last 150 s of the recording, as revealed by the mean‐centered PLS analysis. (a) An overall contrast across the conditions and the corresponding map of z‐scores, each associated with a unique combination of a frequency and brain region. The contrast on the left subplot differentiates the first 150 s of NH versus the last 150 s of NH and the first and last 150 s of NN and HH (p < .001). Large in magnitude negative z‐scores (shown in blue) robustly support the overall contrast, indicating lower (and in this case, more negative) correlations between EEG absolute power and SpO2 in the NH condition. (b) The same map of z‐scores, as in (a), but with the z‐scores shown on the brain template. The alpha rhythm exhibits the greatest effect due the first 150 s of the NH condition. This suggests that the global inhibition or a global idle brain state due to the alpha rhythm is due to the active desaturation occurring during the first 150 s of NH. HH Beg, first 150 s of hypobaric hypoxia; HH End, last 150 s of hypobaric hypoxia, NH Beg, first 150 s of…

• 

  FIGURE 11. Differences in the correlations between EEG relative power and SpO2 across the three conditions at the group level split into the first and last 150 s of the recording, as revealed by the mean‐centered PLS analysis. (a) An overall contrast across the conditions and the corresponding map of z‐scores, each associated with a unique combination of a frequency and brain region. The contrast on the left subplot differentiates the first 150 s of NH versus the last 150 s of NH and the first and last 150 s of NN and HH (p = .001). Large in magnitude negative z‐scores (shown in blue) robustly support the overall contrast, indicating lower (and in this case, more negative) correlations between EEG relative power and SpO2 in the NH condition. Positive z‐scores (shown in red) indicate higher and more positive correlations between EEG relative power and SpO2. (b) The same map of z‐scores, as in (a), but with the z‐scores shown on the brain template. A shift in power is observed toward the alpha band and away from the slower frequency rhythms. This shift in power is observed during the active desaturation phase of the first 150 s of NH. HH Beg, first 150 s of hypobaric hypoxia; HH End, last 150…

• 

  FIGURE 12. Differences in the correlations between MSE and SpO2 across the three conditions at the group level split into the first and last 150 s of the recording, as revealed by the mean‐centered PLS analysis. (a) An overall contrast across the conditions and the corresponding map of z‐scores, each associated with a unique combination of a frequency and brain region. The contrast on the left subplot differentiates the first 150 s of NH and the negligible first 150 s of NN versus the last 150 s of NH and NN, and the first and last 150 s of HH (p = .014). Large in magnitude positive z‐scores (shown in red) robustly support the overall contrast, indicating higher (and in this case, more positive) correlations between MSE and SpO2 in the first 150 s of the NH condition. Negative z‐scores (shown in blue) indicate lower and more negative correlations between MSE and SpO2. (b) The same map of z‐scores, as in (a), but with the z‐scores shown on the brain template. The MSE results seem to show a distinct spatiotemporal pattern of correlations, with a clear distinction of the posterior section of the cortex exhibiting positive correlations occurring during the first 150 s of the NH condition. At…