A deep neural network and transfer learning combined method for cross-task classification of error-related potentials

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Error-related potentials (ErrPs) are brain responses to errors and unexpected events.
• Accurate ErrP detection is crucial for Brain-Computer Interfaces (BCIs).
• Current methods struggle with EEG signal non-stationarity and limited datasets.

Purpose of the Study:

• To develop a deep learning model for accurate ErrP classification.
• To address challenges of limited ErrP datasets and signal non-stationarity.
• To improve generalization of ErrP detection across different tasks and sessions.

Main Methods:

• A deep learning model combining convolutional layers and transformer encoders was developed.
• Transfer learning was employed for effective model training.
• Publicly available datasets were utilized for model evaluation.

Main Results:

• Achieved 78% average accuracy in cross-task classification, outperforming baselines.
• Demonstrated superior performance in leave-one-subject-out (71.81%), within-session (78.74%), and cross-session (77.01%) scenarios.
• The model effectively reduces the need for extensive task-specific training data.

Conclusions:

• The proposed deep learning approach enhances ErrP detection accuracy and generalization.
• This method mitigates challenges associated with limited ErrP datasets.
• The findings provide a foundation for future research in ErrPs and BCI applications.