On the Impact of Proprioception in EEG Representations and Decoding During Human-Hand Exoskeleton Interaction.

Area of Science:

• Neuroscience
• Robotics
• Biomedical Engineering

Background:

• Brain-computer interfaces (BCI) using electroencephalogram (EEG) offer potential for hand exoskeleton control in motor augmentation and neurorehabilitation.
• Understanding the role of proprioception from exoskeleton interaction is crucial for natural BCI control.

Purpose of the Study:

• To quantify EEG representations and BCI performance during hand exoskeleton use.
• To investigate the impact of proprioceptive feedback on sensorimotor EEG activity and BCI classification accuracy.

Main Methods:

• Monitored full-scalp EEG in 25 healthy subjects performing imagined (IM), passive (PM), and congruent imagined and passive (IPM) finger flexion tasks with a cable-driven hand exoskeleton.
• Analyzed alpha and beta band power changes in the sensorimotor area.
• Utilized machine learning models to classify movement conditions (IPM vs. REST, PM vs. REST, IPM vs. PM, IM vs. REST).

Main Results:

• Significantly stronger alpha and beta band desynchronization in sensorimotor areas for PM and IPM tasks compared to IM.
• High accuracy in classifying exoskeleton-assisted movements from rest (IPM vs. REST: 0.80 ± 0.07; PM vs. REST: 0.72 ± 0.10), with IPM showing the highest accuracy.
• Lower accuracy in distinguishing between IPM and PM (0.61 ± 0.09) compared to detecting motor intention without the exoskeleton (IM vs. REST: 0.73 ± 0.08).

Conclusions:

• Sensorimotor EEG activity effectively tracks proprioceptive feedback from hand exoskeletons.
• While proprioceptive feedback is detectable, accurately detecting motor intention during exoskeleton use remains a significant challenge.
• Development of advanced decoders and control strategies is essential for continuous BCI-actuated hand exoskeleton systems.