Classification of Resting-State Status Based on Sample Entropy and Power Spectrum of Electroencephalography (EEG).

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalogram (EEG) is crucial for diagnosing brain disorders and monitoring brain dynamics in healthy individuals.
• Understanding resting-state brain activity, specifically the differences between eyes open (EO) and eyes closed (EC) conditions, is vital for neurological research.
• EEG serves as a key interface for studying brain states and their modulation by external conditions.

Purpose of the Study:

• To investigate the efficacy of EEG signal analysis in discriminating between resting brain states with eyes open (EO) and eyes closed (EC).
• To compare the performance of various feature extraction techniques (frequency bands and entropy) and machine learning classifiers for classifying EO and EC states.
• To identify the most accurate methods for distinguishing between these two fundamental resting brain conditions using EEG data.

Main Methods:

• Utilized sixteen-channel EEG data recorded during resting states with eyes open and eyes closed.
• Extracted features using Fast Fourier Transform (FFT) for conventional frequency bands and Sample Entropy (SE) for signal complexity.
• Employed six machine learning classifiers: Logistic Regression (LR), K-Nearest Neighbors (KNN), Linear Discriminant (LD), Decision Tree (DT), Support Vector Machine (SVM), and Gaussian Naive Bayes (GNB) to classify the states.

Main Results:

• Logistic Regression (LR) and Support Vector Machine (SVM) achieved the highest classification accuracy of 97% when using conventional frequency bands.
• Linear Discriminant (LD), KNN, and DT classifiers showed accuracies of 95%, 93%, and 92%, respectively, with frequency band features.
• Sample Entropy (SE) as a feature resulted in slightly lower, but still high, classification accuracies, with SVM and LD performing best at 92% and 90%.

Conclusions:

• EEG signal analysis, particularly using frequency band features, is highly effective in differentiating between eyes open and eyes closed resting brain states.
• Logistic Regression and Support Vector Machine are robust classifiers for this specific brain state discrimination task, offering superior accuracy.
• The study highlights the potential of EEG feature extraction and machine learning for objective assessment of resting-state brain dynamics.