Rehabilitation Based on BCI: An Innovative Enhancement for Sensorimotor Cortex Rhythms Systemization

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Real-time electroencephalogram (EEG) analysis is crucial for brain-computer interfaces (BCIs) in rehabilitation.
• Nonstationarity in EEG signals presents a significant challenge for accurate data interpretation.
• Existing methods often struggle to effectively capture multi-domain characteristics of EEG for motor imagery tasks.

Purpose of the Study:

• To develop a novel strategy for categorizing EEG signals in real-time BCIs for rehabilitation.
• To enhance the accuracy of predicting Sensory-Motor Rhythms (SMR) by addressing EEG nonstationarity.
• To create a robust systemization model for motor movement and imagery classification.

Main Methods:

• Utilized Five Cross-Common Spatial Patterns (FCCSP) for extracting multi-domain EEG characteristics.
• Implemented a Butterworth bandpass filter optimized by FCCSP to identify optimal bandwidth for beta waves.
• Integrated the FCCSP method with a Hybrid Systemization of Correlated Feature Removal classifier for improved predictive models.

Main Results:

• Achieved a significant accuracy increase from 57.14% to 85.71% for real-time SMR prediction after preprocessing.
• Reached an accuracy of 97.94% on public domain SMR data, demonstrating robust performance.
• Validated the effectiveness of the FCCSP strategy in enhancing SMR prediction efficiency on both real-time and PhysioNet datasets.

Conclusions:

• The proposed FCCSP-based strategy effectively mitigates EEG nonstationarity for improved BCI performance.
• The developed system demonstrates high accuracy and robustness in real-time SMR prediction for rehabilitation applications.
• This approach offers a promising advancement for enhancing the efficacy of BCIs in clinical settings.