Spatio-temporal deep learning for EEG-fNIRS brain computer interface.

Area of Science:

• Neuroscience and Biomedical Engineering
• Signal Processing

Background:

• Motor imagery brain signal classification is crucial for brain-computer interfaces (BCIs).
• Traditional methods often focus solely on temporal signal dynamics, potentially overlooking valuable spatial information from electrode placement.

Purpose of the Study:

• To improve motor imagery brain signal classification accuracy.
• To investigate the benefits of incorporating both temporal and spatial information from electroencephalography (EEG) and functional near-infrared spectroscopy (fNIRS) data.

Main Methods:

• A deep neural network architecture was developed to process multi-modal brain data.
• The model simultaneously utilized temporal signal features and spatial information derived from electrode locations.
• Both motor-imagery EEG and bi-modal EEG-fNIRS datasets were employed for training and validation.

Main Results:

• The proposed method demonstrated superior performance compared to approaches using only temporal data.
• Integrating spatial information from electrode configurations significantly enhanced classification accuracy.
• Results were validated across different experimental scenarios and comparative methods.

Conclusions:

• Combining temporal and spatial information is a highly effective strategy for motor imagery classification.
• Deep learning models can successfully leverage multi-modal brain data for improved BCI performance.
• This approach holds significant promise for advancing the capabilities of brain-computer interfaces.