A Generic Transferable EEG Decoder for Online Detection of Error Potential in Target Selection

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Brain-computer interfaces (BCI) offer rehabilitative potential but are prone to errors.
• Error Related Potentials (ErrP) are crucial EEG signals for detecting system inaccuracies.
• Effective online ErrP detection is vital for closed-loop BCI systems.

Purpose of the Study:

• To propose a novel scheme for online detection of error feedback directly from EEG signals.
• To develop a transferable error detection system applicable across different sessions and subjects.
• To enhance the reliability and real-time feedback capabilities of BCI systems.

Main Methods:

• Utilized a P300-speller dataset for training and testing.
• Developed a decoder using an ensemble of linear discriminant analysis, quadratic discriminant analysis, and logistic regression classifiers.
• Trained the decoder on EEG features from 16 subjects and tested on 10 independent subjects for single-trial classification.

Main Results:

• Achieved an accuracy of 73.97% in detecting ErrP signals.
• Obtained an F1-score of 83.53%, indicating robust performance.
• Reported an Area Under the Curve (AUC) of 73.18% for the classification model.

Conclusions:

• The proposed scheme demonstrates effective online detection of ErrP signals in a transferable manner.
• The developed BCI system shows promise for real-time error correction in rehabilitative applications.
• The ensemble classifier approach provides a reliable method for error detection across subjects and sessions.