Reduce brain computer interface inefficiency by combining sensory motor rhythm and movement-related cortical...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Brain Computer Interface (BCI) systems enable communication and control for individuals with severe motor impairments.
• BCI inefficiency, affecting 10-50% of users, is often attributed to limitations in Sensory Motor Rhythm (SMR) features.
• Previous studies primarily focused on SMR, leaving the potential of other features underexplored for addressing BCI inefficiency.

Purpose of the Study:

• To investigate the effectiveness of combining SMR and Movement-Related Cortical Potential (MRCP) features in mitigating BCI inefficiency.
• To assess the occurrence of BCI inefficiency using both SMR and MRCP features across a large user group.
• To determine if feature combination can improve classification accuracy for users who perform poorly with SMR alone.

Main Methods:

• Recorded electroencephalogram (EEG) data from 93 subjects over two sessions, including resting states and movement tasks.
• Extracted SMR and MRCP features using Common Spatial Pattern (CSP) and template matching, respectively.
• Employed a winner-take-all strategy combining posterior probabilities from Linear Discriminant Analysis to integrate SMR and MRCP features for pattern recognition.

Main Results:

• SMR and MRCP features demonstrated high complementarity with weak intercorrelation.
• In subjects with SMR accuracies below 70%, combining SMR and MRCP features improved average accuracy from 62% to 79%.
• The accuracies achieved through feature combination surpassed the defined inefficiency threshold.

Conclusions:

• The combination of SMR and MRCP features offers a practical strategy to reduce BCI inefficiency.
• MRCP features provide comparable classification accuracy across user groups with both poor and good SMR performance.
• This study suggests that 'BCI inefficiency' may be more accurately termed 'SMR inefficiency', highlighting the critical role of SMR limitations.