Classification of propofol-induced sedation states using brain connectivity analysis

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain connectivity analysis provides insights into neural processes.
• Propofol is an anesthetic agent affecting brain states.
• Electroencephalography (EEG) is a common tool for measuring brain activity.

Purpose of the Study:

• To evaluate the performance of binary classification for Propofol-induced sedation states.
• To investigate the utility of partial Granger causality analysis for brain connectivity.
• To assess classification accuracy using various machine learning models.

Main Methods:

• Utilized EEG signals from a database with four sedation states: baseline, mild, moderate, and recovery.
• Applied partial Granger causality analysis to derive brain connectivity measurements.
• Evaluated five classifiers: k-nearest neighbor, support vector machine, linear discriminant analysis, Bayesian discriminant analysis, and extreme learning machine.
• Focused on eight EEG sensors and short signal segments (4 seconds).

Main Results:

• Achieved an Area Under the ROC Curve (AUC) of approximately 0.75 for classifying sedation states.
• Demonstrated that different Propofol-induced sedation states can be identified with this approach.
• Highlighted the effectiveness of short EEG signal segments (4 seconds).

Conclusions:

• Scalp-level connectivity measures possess significant discriminant power for online brain monitoring.
• Partial Granger causality analysis is a viable method for assessing Propofol-induced sedation states.
• The findings support the potential for real-time monitoring of anesthetic effects using EEG connectivity.