Feature domain-specific movement intention detection for stroke rehabilitation with brain-computer interfaces

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Brain-computer interfaces (BCI) utilize electrical stimulation for neuromodulation in stroke rehabilitation.
• Pairing movement intentions with somatosensory feedback is a key strategy.
• Detecting movement intentions using temporal and spectral features from EEG has yielded conflicting results.

Purpose of the Study:

• To investigate the efficacy of complexity measures for movement intention detection in single-trial EEG.
• To compare the detection performance of temporal, spectral, and complexity features.
• To evaluate BCI performance for stroke rehabilitation.

Main Methods:

• EEG data were recorded from 39 healthy subjects and 11 stroke patients performing or imagining isometric hand grasps and dorsiflexions.
• EEG signals were pre-processed into Background EEG and movement intention epochs.
• Temporal, spectral, and complexity features were extracted, reduced using sequential forward selection, and classified using linear discriminant analysis.

Main Results:

• Accuracies of 82-87% for foot movement intention and 74-80% for hand movement intention were achieved.
• Temporal features were most effective for foot movement detection, while spectral features were superior for hand movements.
• Complexity features demonstrated lower detection performance compared to temporal and spectral features.

Conclusions:

• Temporal and spectral features are effective for movement intention detection in BCI applications for stroke rehabilitation.
• Complexity measures show potential but are currently less effective than traditional features.
• This research contributes to optimizing BCI-driven neuromodulation strategies for improved motor recovery.