A functional source separation algorithm to enhance error-related potentials monitoring in noninvasive brain-computer...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Error-related Potentials (ErrPs) are measurable EEG signals reflecting error detection.
• ErrPs are crucial for Brain-Computer Interfaces (BCIs) in error correction and adaptation.
• Current methods for ErrP detection face limitations in accuracy and efficiency.

Purpose of the Study:

• To propose a novel semi-supervised algorithm, Functional Source Separation (FSS), for estimating spatial filters.
• To enhance the detection and classification of ErrPs from EEG data.
• To improve the performance of BCIs by increasing ErrP detection accuracy.

Main Methods:

• EEG data from six subjects were analyzed.
• The proposed FSS algorithm was compared against the xDAWN algorithm and single channels (Cz, FCz).
• Single-trial classification performance was evaluated using a Bayesian Linear Discriminant Analysis (BLDA) classifier.

Main Results:

• The FSS-based method achieved a classification accuracy of 0.92, sensitivity of 0.95, specificity of 0.81, and F1-score of 0.95.
• These results significantly outperformed single channels (Cz, FCz) and the xDAWN algorithm.
• FSS demonstrated superior performance in single-trial classification of ErrPs.

Conclusions:

• The FSS-based method significantly enhances the single-trial detection accuracy of ErrPs.
• FSS provides a more effective spatial filter for ErrP detection compared to existing methods.
• This advancement holds promise for more robust and accurate BCIs.