Related Experiment Video
Updated: Nov 29, 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
1.8K
An EEG channel selection method for motor imagery based brain-computer interface and neurofeedback using Granger
Hesam Varsehi1, S Mohammad P Firoozabadi2
1Department of Biomedical Engineering, Tarbiat Modares University, Tehran, Iran.
Summary
This study introduces a novel Granger causality (GC) method for electroencephalogram (EEG) channel selection in motor imagery (MI) brain-computer interfaces (BCI). The approach enhances BCI performance by reducing channels while improving classification accuracy.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Machine Learning
Background:
- Motor imagery (MI) brain-computer interfaces (BCI) and neurofeedback (NF) utilize electroencephalogram (EEG) signals for motor function enhancement and neurological restoration.
- Effective BCI and NF implementations often require channel selection to reduce noisy and redundant EEG data.
- Existing channel selection methods may not fully leverage the underlying causal interactions within EEG signals during MI tasks.
Purpose of the Study:
- To propose a novel EEG channel selection method based on Granger causality (GC) analysis for improved MI-BCI and NF performance.
- To investigate the effectiveness of causal interactions in reducing EEG channels while maintaining or enhancing classification accuracy.
- To integrate machine learning for artifact detection and optimize feature extraction and classification for MI tasks.
Main Methods:
- A novel channel selection method employing Granger causality (GC) analysis to identify causally interacting EEG channels.
- Machine learning-based clustering of Independent Component Analysis (ICA) components to differentiate between artifact and normal EEG signals.
- Feature extraction using Common Spatial Pattern (CSP) and regularized CSP (RCSP) followed by classification with k-NN, SVM, and LDA.
Main Results:
- The proposed GC-based channel selection method achieved high classification performance with only eight selected EEG channels.
- The method resulted in 93.03% accuracy, 92.93% sensitivity, and 93.12% specificity on the Physionet MI dataset.
- Performance metrics showed significant improvements (3.95-4.13%) compared to a correlation-based channel selection method.
Conclusions:
- Granger causality analysis offers an effective approach for EEG channel selection in MI-BCI and NF applications.
- Causal constraint-based channel selection can lead to a reduced number of channels with superior classification performance.
- The developed method provides a promising strategy for optimizing BCI and NF systems.
