Classifying Multi-Level Stress Responses From Brain Cortical EEG in Nurses and Non-Health Professionals Using Machine...

Area of Science:

• Neuroscience
• Psychiatry
• Biomedical Engineering

Background:

• Mental stress is a significant societal issue and a key focus in psychiatric research.
• Identifying reliable bio-markers for stress detection and prediction is crucial.
• Electroencephalograms (EEGs) are utilized for bio-marker identification, but traditional methods struggle with multiple stress conditions.

Purpose of the Study:

• To enhance the accuracy of EEG-based stress detection, particularly for multiple stress conditions.
• To investigate the efficacy of latent-based representations of bio-markers in improving stress prediction.
• To compare the performance of traditional EEG bio-markers with their latent representations.

Main Methods:

• Evaluation of three common EEG bio-markers: brain load index (BLI), spectral power values (alpha, beta, theta bands), and relative gamma (RG).
• Utilized latent representations of these bio-markers.
• Employed four standard classifiers to assess performance.

Main Results:

• Spectral power value bio-markers achieved 83% accuracy in stress prediction.
• Latent representations of spectral power values significantly improved accuracy to 91%.
• Latent-based approaches demonstrated superior performance over traditional bio-markers.

Conclusions:

• Latent-based representations offer a substantial improvement for EEG-based stress detection.
• This methodology shows promise for more accurate prediction of multiple stress conditions.
• The findings suggest a new direction for developing advanced bio-markers for mental stress.