Assessing Cumulative Mental Fatigue via EEG-Based Machine Learning in a Multiday High-Intensity Contest

Area of Science:

• Neuroscience
• Machine Learning
• Occupational Health

Background:

• Cumulative mental fatigue is a significant workplace hazard impacting safety and productivity.
• Developing objective measures for fatigue detection is crucial for risk management.

Purpose of the Study:

• To create a machine learning framework for detecting mental fatigue using optimized resting-state electroencephalography (rs-EEG) features.
• To validate a high-stress cognitive competition paradigm for inducing and studying fatigue.

Main Methods:

• EEG signals were recorded under eyes-closed (EC) and eyes-open (EO) conditions.
• 544 features were extracted, and Support Vector Machine Recursive Feature Elimination (SVM-RFE) was used for selection.
• The model index (MMR) was correlated with the Stanford Sleepiness Scale (SSS) and sleep duration.

Main Results:

• A subset of 65 EC EEG features achieved 90.37% accuracy in classifying fatigue states.
• The EC model outperformed the EO model (86.54% accuracy).
• MMR showed significant negative correlation with SSS and positive correlation with sleep duration.

Conclusions:

• Eyes-closed resting-state EEG is highly effective for monitoring cumulative mental fatigue.
• A quantifiable relationship between EEG markers and sleep was established.
• The framework is practically feasible for occupational fatigue risk assessment.