Enhancing EEG-based cross-day mental workload classification using periodic component of power spectrum

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Day-to-day variability in electroencephalogram (EEG) signals presents a major obstacle for EEG-based passive brain-computer interfaces (pBCIs).
• Current machine learning models require time-consuming daily recalibration, limiting real-world applications like mental workload monitoring.
• This variability hinders the consistent decoding of brain activity.

Purpose of the Study:

• To investigate the day-to-day stability of raw power spectral density (PSD) and its decomposed periodic and aperiodic components.
• To assess the feasibility of using these periodic components for improved cross-day mental workload classification.
• To enhance the robustness of pBCIs for practical applications.

Main Methods:

• Decomposition of raw EEG power spectral density (PSD) into periodic and aperiodic components using the Fitting Oscillations and One-Over-F algorithm.
• Analysis of the day-to-day stability of these components.
• Validation of cross-day mental workload classification performance using periodic components.

Main Results:

• The periodic component of EEG exhibited superior day-to-day stability compared to raw PSD and the aperiodic component.
• Classification accuracy for mental workload using periodic components reached 84.2% ± 11.0%, significantly outperforming raw PSD (69.9% ± 18.5%) and the aperiodic component (69.4% ± 19.2%).
• Periodic components demonstrated better cross-day classification performance.

Conclusions:

• Periodic components of EEG possess inherent day-to-day stability, making them promising for brain state decoding.
• Utilizing periodic components can significantly improve cross-day classification accuracy in pBCIs.
• These findings pave the way for more robust and practical pBCI applications without constant recalibration.