Non-invasive decoding of hand movements from electroencephalography based on a hierarchical linear regression model

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Non-invasive brain-computer interfaces (BCIs) use electroencephalography (EEG) to translate brain signals into device commands.
• Current BCI decoding efficiency is limited by an incomplete understanding of effective decoding parameters.
• Restoring hand function post-stroke is a significant challenge for assistive technologies.

Purpose of the Study:

• To investigate the influence of brain functional network parameters on BCI decoding efficiency.
• To identify optimal EEG channels and frequency bands for improved decoding performance.
• To evaluate a hierarchical linear regression (HLR) model for decoding upper-limb movements.

Main Methods:

• Five subjects performed self-initiated upper-limb movements across three experimental phases.
• A hierarchical linear regression (HLR) model was developed to analyze decoding efficiency based on brain network characteristics.
• The Kruskal-Wallis test was used to select optimal EEG channels and frequency bands.

Main Results:

• The HLR model achieved a Pearson correlation coefficient (R) of 0.66 for decoding movement trajectories, outperforming the multiple linear regression model (R=0.46).
• Optimal EEG channels and sensitive frequency bands were identified using statistical analysis.
• Successful decoding of free and conical helix upper-limb movements was demonstrated.

Conclusions:

• The HLR model significantly enhances decoding efficiency in EEG-based BCIs.
• This approach shows potential for aiding hand movement restoration in post-stroke rehabilitation.
• Further development of EEG-based BCIs using HLR could improve assistive technology capabilities.