A Hybrid Covert Attention-Augmented Motor Imagery Paradigm for Brain-Computer Interfaces.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Human-Computer Interaction

Background:

• Motor imagery (MI) is crucial for brain-computer interfaces (BCIs) and neurorehabilitation, but weak bio-markers from inexperienced users limit effectiveness.
• Current BCIs struggle with reliable classification due to insufficient or weak neural signals during MI tasks.
• Enhancing MI with additional cognitive markers can improve feature representation and feedback reliability for better training.

Purpose of the Study:

• To propose a novel covert attention-augmented motor imagery (CAA-MI) paradigm to enrich neural features for BCI decoding.
• To develop a transformer-based multi-branch EEG fusion network (TMEF-Net) for decoding CAA-MI signals.
• To evaluate the performance of CAA-MI against traditional MI (T-MI) in improving BCI accuracy and neurorehabilitation potential.

Main Methods:

• Developed a hybrid BCI approach integrating covert spatial attention (CSA) with motor imagery (MI).
• Utilized a transformer-based multi-branch EEG fusion network (TMEF-Net) for signal processing and classification.
• Conducted experiments with 17 subjects comparing CAA-MI and T-MI under varying decoding window lengths.

Main Results:

• CAA-MI demonstrated superior performance over T-MI in both intra-subject (89% vs. 82%) and inter-subject (81% vs. 76%) evaluations with a 3-s decoding window.
• The CAA-MI paradigm showed significant improvements, especially within shorter decoding windows.
• Neurophysiological analysis indicated broader brain activation patterns with CAA-MI, particularly in occipital-parietal and sensorimotor regions, yielding more discriminative EEG features.

Conclusions:

• The proposed CAA-MI paradigm effectively augments MI with covert spatial attention, generating richer and more discriminative neural features.
• CAA-MI offers a promising strategy for enhancing the robustness and usability of neurorehabilitation-oriented BCIs.
• The TMEF-Net architecture successfully decodes the complex EEG signals elicited by the CAA-MI paradigm, outperforming traditional MI approaches.