Non-linear online low-frequency EEG decoding of arm movements during a pursuit tracking task

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Decoding upper-limb movements from invasive recordings is established, but non-invasive methods face challenges.
• Linear decoders for electroencephalogram (EEG) have decoded movement but with reduced signal amplitude, impacting user experience.
• Previous work introduced a non-linear decoder combining directional and non-directional movement information for offline analysis.

Purpose of the Study:

• To evaluate the efficacy of a non-linear decoder for online, real-time decoding of upper-limb movements using EEG.
• To assess the potential of this non-linear decoder to improve feedback in brain-computer interfaces (BCIs).
• To determine if speed information is encoded in EEG signals for real-time movement decoding.

Main Methods:

• Five healthy participants controlled a robotic arm, with control transitioning from their hand to EEG.
• A non-linear decoder was implemented for online processing of EEG data during a target-tracking task.
• Correlations between actual and decoded movements were calculated to assess decoding accuracy.

Main Results:

• Decoding of upper-limb movements using the non-linear EEG decoder generally showed correlations above chance level.
• The results indicate that information related to movement speed is encoded within EEG signals.
• The non-linear decoder proved suitable for decoding arm movements in real-time.

Conclusions:

• The developed non-linear decoder is effective for online EEG-based decoding of upper-limb movements.
• This approach holds promise for enhancing real-time feedback in BCIs for robotic arm control.
• The findings suggest that EEG contains encoded information about movement speed, crucial for accurate decoding.