Cortical-SSM: A deep state space model for motor imagery decoding from EEG signals.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Machine Learning

Background:

• Electroencephalogram (EEG) signal classification is crucial for brain-computer interfaces (BCIs) in motor rehabilitation.
• Physiological artifacts and limited dependency capture challenge current Transformer-based EEG classification methods.
• Accurate decoding of motor imagery (MI) signals is vital for assistive technologies.

Purpose of the Study:

• To introduce Cortical-SSM, a novel deep state space model architecture for enhanced EEG signal analysis.
• To improve the classification accuracy and interpretability of motor imagery (MI) EEG signals.
• To overcome limitations of Transformer models in capturing fine-grained EEG signal dependencies.

Main Methods:

• Developed Cortical-SSM, an architecture extending deep state space models.
• Integrated temporal, spatial, and frequency domain dependencies within EEG signals.
• Validated the model on two large-scale public MI EEG datasets (>50 subjects).

Main Results:

• Cortical-SSM significantly outperformed baseline methods on benchmark datasets.
• Model explanations confirmed the capture of neurophysiologically relevant EEG signal features.
• Demonstrated robust and interpretable performance for MI EEG decoding.

Conclusions:

• Cortical-SSM offers a reliable, interpretable alternative to attention-based models for MI EEG.
• Physiologically grounded feature learning enhances subject-independent EEG classification.
• The method supports the development of practical, clinically deployable BCI systems.