Effects of Cognitive Distraction on Upper Limb Movement Decoding From EEG Signals.

Area of Science:

• Neuroscience and Biomedical Engineering
• Brain-Computer Interfaces (BCI)
• Rehabilitation Technology

Background:

• Decoding hand movements from electroencephalograms (EEG) is crucial for assisting upper limb-impaired patients.
• Real-life application of EEG-based systems is hindered by the lack of consideration for cognitive distraction in existing decoding methods.
• Investigating the impact of cognitive distraction is essential for developing practical and effective assistive technologies.

Purpose of the Study:

• To investigate the effects of cognitive distraction on the performance of hand movement decoding from EEG signals.
• To propose and evaluate a robust decoding method capable of mitigating the impact of cognitive distraction.

Main Methods:

• Proposed a novel decoding method (RM-GNBC) utilizing Riemannian Manifold for affine invariant feature extraction and a Gaussian Naive Bayes classifier.
• Compared the proposed RM-GNBC method with Tangent Space Linear Discriminant Analysis (TSLDA) and a baseline method.
• Evaluated decoding performance using both experimental and simulated EEG data under conditions with and without cognitive distraction.

Main Results:

• Riemannian-based methods (RM-GNBC and TSLDA) demonstrated higher accuracy and smaller performance decreases under cognitive distraction compared to the baseline.
• RM-GNBC achieved 6% higher accuracy than TSLDA without distraction (p=0.026) and 5% higher accuracy with distraction (p=0.137).
• Results indicate superior robustness of Riemannian-based approaches to cognitive distraction in EEG-based hand movement decoding.

Conclusions:

• Riemannian-based methods offer enhanced robustness against cognitive distraction for EEG-based hand movement decoding.
• The developed RM-GNBC method shows significant potential for improving brain-computer interfaces (BCI) in real-life assistive applications.
• This study opens avenues for exploring distraction effects in other BCI paradigms.