Improved Cognitive Vigilance Assessment after Artifact Reduction with Wavelet Independent Component Analysis

Area of Science:

• Neuroscience
• Signal Processing
• Human Factors Engineering

Background:

• Accurate vigilance level assessment is crucial in critical environments to prevent human error.
• Electroencephalogram (EEG) is a key modality for vigilance monitoring, but susceptible to artifacts.
• Artifacts in EEG signals, from eye movements or muscle activity, hinder precise vigilance assessment.

Purpose of the Study:

• To evaluate the efficacy of wavelet Independent Component Analysis (wICA) for reducing EEG artifacts.
• To compare the performance of wICA against traditional Independent Component Analysis (ICA) in vigilance assessment.
• To identify brain regions most affected by vigilance decrement using topographical analysis.

Main Methods:

• An experiment involving nine subjects to induce alert and vigilance decrement states using the Stroop Color-Word Test.
• Application of both ICA and wICA methods for preprocessing EEG data.
• Utilizing five different classifiers to analyze feature extraction performance.
• Comparison of power spectral density changes across brain regions using topographical maps.

Main Results:

• wICA preprocessing significantly outperformed ICA in feature extraction for vigilance assessment.
• Mean classification accuracy improved from 84.66% (ICA) to 96.9% (wICA) in delta, theta, and alpha bands.
• Topographical analysis indicated frontal and central brain regions are most sensitive to vigilance decrement.
• No significant improvement was observed in the beta band using wICA compared to ICA.

Conclusions:

• Wavelet ICA (wICA) offers a superior alternative for EEG artifact reduction in vigilance assessment compared to traditional ICA.
• The wICA method enhances the accuracy of distinguishing between alert and vigilance decrement states.
• Frontal and central brain regions show the most significant changes in power spectral density during vigilance decrement.