A finer-grained high altitude EEG dataset for hypoxia levels assessment

Figures

• 

  Fig. 1. A schematic diagram depicting the hypobaric chamber for high-altitude simulation, the physiological signal acquisition device, and the airflow direction.

• 

  Fig. 2. Altitude gradient adjustment scheme. The blue dashed lines indicate the altitude adjustment protocol during the experiment. The blue solid lines represent periods of stable altitude, during which EEG data was collected. Red triangles mark the start and end times of the data collection process.

• 

  Fig. 3. The scheme for labeling the collected EEG data based on oxygen saturation levels.

• 

  Fig. 4. The structure of the EEG dataset in BIDS format.

• 

  Fig. 5. The EEG data quality distribution across channels and altitudes for subject 09. The quality distribution chart illustrates the quality of EEG data collected from subject Sub09 during the experiment. The horizontal axis represents altitude groups, while the vertical axis lists all the EEG electrodes. The red blocks denote bad channels, whereas the white blocks indicate good ones.

• 

  Fig. 6. The distribution of epoch counts and PSD across different levels of hypoxia. (a) The bar chart presents the distribution of epoch counts at various hypoxia levels in the preprocessed data. The scatter plot illustrates the distribution of epochs for each volunteer across different hypoxia levels. (b) represents the post-hoc Tukey HSD Test results for PSD across four hypoxia levels. Perform post-hoc multiple comparisons Tukey HSD test on the average PSD under different levels. The level of significance is indicated by asterisks: * for p < 0.05, ** for p < 0.01, and *** for p < 0.001.

• 

  Fig. 7. The binary classification test results of SVM and EEGNet. (a) and (b) illustrate the classification accuracies for a binary-class problem using SVM and EEGNet. Each subject’s classification performance is represented by the blue bars, while the black dashed line indicates the mean classification accuracy across all subjects. (c) and (d) display the mean confusion matrices from a 10-fold cross-validation for the binary-class hypoxia levels classification, using SVM and EEGNet.

• 

  Fig. 8. The 4-category classification test results of SVM and EEGNet. (a) and (b) illustrate the classification accuracies for a 4-category problem using SVM and EEGNet. Each subject’s classification performance is represented by the blue bars, while the black dashed line indicates the mean classification accuracy across all subjects. (c) and (d) display the mean confusion matrices from a 10-fold cross-validation for the 4-category hypoxia levels classification, using SVM and EEGNet.