A Personalized Predictor of Motor Imagery Ability Based on Multi-frequency EEG Features

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interfaces (BCIs) leverage motor imagery (MI) to decode brain intentions for enhanced control.
• Significant intra-individual variability in MI ability hinders current BCI system adaptability and training efficacy.
• Accurate prediction of MI ability is crucial for optimizing BCI performance and user training.

Purpose of the Study:

• To develop and validate a novel predictor for motor imagery (MI) ability using multi-frequency electroencephalography (EEG) features.
• To assess the predictor's performance across different MI paradigms, including video-guided and traditional approaches.
• To establish a foundation for personalized BCI training strategies.

Main Methods:

• A predictor model was developed utilizing multi-frequency EEG features.
• Two distinct MI paradigms were employed for validation: a video-guided paradigm and a traditional MI paradigm.
• The predictor's precision was evaluated across all subjects in both experimental conditions.

Main Results:

• The proposed MI ability predictor achieved high prediction precision, exceeding 85% for all subjects in both paradigms.
• Individual prediction precision reached up to 96% in certain applications.
• The predictor demonstrated consistent accuracy in forecasting MI ability across different states and paradigms.

Conclusions:

• The developed predictor accurately forecasts individuals' motor imagery ability in various states.
• This predictor offers a scientific basis for tailoring BCI training protocols to individual users.
• Implementing this predictor can significantly enhance the effectiveness of motor imagery-based BCI training.