Classification of upper limb center-out reaching tasks by means of EEG-based continuous decoding techniques

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Brain-machine interfaces (BMIs) aim to decode upper limb kinematics from brain signals for prosthetic control.
• Previous studies suggest feasibility using low-frequency electroencephalography (EEG) components, but results require further validation.
• This study investigates decoding upper limb kinematics from EEG during center-out reaching tasks.

Purpose of the Study:

• To assess the feasibility of decoding upper limb kinematics from EEG signals.
• To evaluate the role of proprioceptive feedback in decoding movement.
• To compare the effectiveness of classifying reaching targets versus continuous trajectory decoding.

Main Methods:

• Multidimensional linear regression was employed for decoding arm movement from EEG signals.
• Passive movements were analyzed to assess the influence of proprioceptive sensory feedback.
• Classification of reaching targets was evaluated as an alternative decoding strategy.

Main Results:

• Arm movement decoding from EEG signals was significantly above chance levels.
• EEG slow cortical potentials contain significant information for decoding active center-out movements.
• Classification of reached targets achieved high accuracy, while decoding of passive movements showed low performance, indicating movement execution is key, not just proprioception.

Conclusions:

• Low-frequency EEG bands are crucial for decoding upper limb kinematics.
• Decoding active movements from EEG is feasible and linked to cortical motor area activation.
• Classifying reached targets is a potentially more suitable real-time BMI methodology than direct hand position decoding.