Improving motor imagery classification during induced motor perturbations.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Motor imagery (MI) is crucial for brain-computer interfaces (BCIs), simulating movements to modulate brain activity.
• Current BCIs face challenges with performance robustness due to involuntary movements, often caused by peripheral stimulation.
• This research addresses the need to improve BCI system reliability under movement-induced perturbations.

Purpose of the Study:

• To test and enhance the robustness of motor imagery-based BCIs against artificially generated limb movements.
• To investigate the performance decrease in BCIs caused by movement perturbations.
• To develop computational strategies for mitigating accuracy drops in BCIs during motor imagery.

Main Methods:

• Conducted BCI sessions with ten participants performing motor imagery of three limbs.
• Introduced neuromuscular stimulation to induce limb movements during specific trials.
• Analyzed 2-class motor imagery classifications with and without induced movement perturbations.
• Applied spatial filtering techniques to reduce neural noise from stimulation.

Main Results:

• BCI performance remained similar to control conditions when induced movements did not involve the imagined limb.
• Spatial filtering significantly alleviated performance drops when induced movements affected the imagined limb.
• Reduced sensorimotor rhythm power correlated with BCI accuracy loss, and residual power predicted user performance under disturbances.

Conclusions:

• Developed methods to ameliorate or eliminate motor-related afferent disturbances in motor imagery tasks.
• Demonstrated that spatial filtering can significantly improve BCI robustness against movement perturbations.
• The findings contribute to enhancing the reliability and practical application of motor imagery-based BCIs.