Classification of EEG based-mental fatigue using principal component analysis and Bayesian neural network

Area of Science:

• Neuroscience
• Computational Intelligence
• Biomedical Engineering

Background:

• Mental fatigue significantly impacts cognitive performance and daily activities.
• Objective detection of mental fatigue is crucial for safety and productivity.
• Electroencephalography (EEG) offers a non-invasive method for monitoring brain activity related to fatigue.

Purpose of the Study:

• To develop and validate an EEG-based classification system for detecting mental fatigue.
• To reduce the dimensionality of EEG data while preserving essential information for fatigue detection.
• To compare the performance of a reduced-dimension EEG analysis with the full-channel analysis.

Main Methods:

• Utilized electroencephalography (EEG) data from 65 healthy participants.
• Applied Principal Component Analysis (PCA) for dimensionality reduction of 26 EEG channels to 6 principal components (PCs).
• Employed Power Spectral Density (PSD) for feature extraction and a Bayesian Neural Network (BNN) for classification.

Main Results:

• Reduced EEG data (6 PCs) retained over 90% of the original information.
• Accurate classification of pre- vs. post-mental load tasks (fatigue detection) was achieved with accuracies of 76% (eyes open) and 75.3% (eyes closed).
• Classification performance using 6 PCs was comparable to using all 26 original EEG channels.

Conclusions:

• Dimensionality reduction using PCA significantly simplifies EEG data analysis for mental fatigue detection.
• The proposed BNN-based system offers an efficient and effective approach to identifying mental fatigue.
• This method has the potential to reduce computational complexity in real-time fatigue monitoring systems.