Functional Connectivity Methods for Multi-Class Mental Workload Classification.

Area of Science:

• Neuroscience
• Cognitive Science
• Biomedical Engineering

Background:

• Network-based features derived from electroencephalography (EEG) signals are increasingly used to classify mental workload.
• Functional connectivity (FC) methods quantify statistical relationships between EEG electrode potentials for feature extraction.

Purpose of the Study:

• To compare the efficacy of three FC-based feature extraction methods—weighted phase lag index (WPLI), imaginary coherence (IC), and layer entanglement (LE)—for classifying mental workload.
• To evaluate the performance of these methods using a support vector machine classifier on data from the Multi-Attribute Task Battery.

Main Methods:

• Comparison of WPLI, IC, and LE for EEG-based mental workload classification.
• Utilized a support vector machine classifier for performance evaluation.
• Tested classification accuracy for both three-level and two-level workload scenarios.

Main Results:

• Layer entanglement (LE) achieved the highest accuracy (89% for three levels, 97% for two levels) when used alone.
• Combined FC methods yielded 88% accuracy for three levels and 97% for two levels.
• Imaginary coherence (IC) and WPLI showed lower accuracies (67% and 61% for three levels, respectively).

Conclusions:

• Layer entanglement (LE) demonstrates superior performance in classifying mental workload from EEG signals compared to IC and WPLI.
• LE-based methods offer potential for accurate mental workload prediction.
• These findings support the development of passive brain-computer interfaces leveraging advanced FC techniques.