Related Experiment Video
Updated: Jun 9, 2025
09:42
Author Spotlight: Using Motor Imagery Brain-Computer Interface to Improve Motor and Cognitive Function in Stroke Patients
Published on: September 1, 2023
1.1K
Compact convolutional transformer for subject-independent motor imagery EEG-based BCIs
Aigerim Keutayeva1, Nail Fakhrutdinov2, Berdakh Abibullaev3
1Institute of Smart Systems and Artificial Intelligence (ISSAI), Nazarbayev University, Astana, 010000, Kazakhstan. aigerim.keutayeva@alumni.nu.edu.kz.
Scientific Reports
|October 29, 2024
Summary
This study introduces EEGCCT, a novel deep learning model for analyzing electroencephalography (EEG) data in brain-computer interfaces (BCIs). EEGCCT improves motor imagery analysis, outperforming existing models with enhanced generalization from limited data.
Area of Science:
- Neuroscience
- Machine Learning
- Biomedical Engineering
Background:
- Electroencephalography (EEG) analysis is vital for brain-computer interfaces (BCIs).
- Challenges in EEG analysis include data complexity and inter-subject variability.
- Limited data generalization is a common issue in EEG datasets.
Purpose of the Study:
- Introduce EEGCCT, a compact convolutional transformer model for motor imagery EEG analysis.
- Enhance generalization capabilities for limited EEG data.
- Improve subject-independent performance in BCIs.
Main Methods:
- Developed EEGCCT, a compact convolutional transformer architecture.
- Validated models on BCI Competition IV datasets 2a and 2b.
- Employed a Leave-One-Subject-Out (LOSO) cross-validation strategy.
Main Results:
- EEGCCT demonstrated superior performance over conventional models like EEGNet.
- Achieved better results than advanced models including Conformer, Hybrid s-CViT, and Hybrid t-CViT.
- Attained 70.12% accuracy with fewer parameters, showcasing enhanced generalization.
Conclusions:
- EEGCCT effectively addresses limitations in EEG data analysis for motor imagery tasks.
- The model offers improved generalization and subject-independent performance.
- EEGCCT represents a significant advancement in BCI development through efficient EEG analysis.
