Improvements in event-related desynchronization and classification performance of motor imagery using instructive dynamic guidance and complex tasks.

BACKGROUND AND

OBJECTIVE

The motor-imagery based brain-computer interface supplies a potential approach for motor-impaired patients, not only to control rehabilitation facilities but also to promote recovery from motor dysfunctions. To improve event-related desynchronization during motor imagery and obtain improved brain-computer interface classification accuracy, we introduce dynamic video guidance and complex motor tasks to the motor imagery paradigm.

METHODS

Eleven participants were included in the experiment; 64-channel electroencephalographic data were collected and analyzed during four motor imagery tasks with different guidance. Time-frequency analysis, spectral-time variation analysis, topographical distribution maps, and statistical analysis were utilized to analyze the event-related desynchronization patterns. Common spatial patterns were used to extract spatial pattern features and support vector machines were used to discriminate the offline classification accuracies in three bands (the alpha band, beta band, alpha and beta band) for comparison.

RESULTS

The experimental outcomes showed that complex motor imagery tasks coupled with dynamic video guidance induced significantly stronger event-related desynchronization than other paradigms, which use simple motor imagery tasks or static guidance. Similar results were obtained during analysis of the motor imagery brain-computer interface classification performance; namely, the highest average classification accuracy in complex and dynamic guidance was improved by approximately 14%, compared with static guidance. For individually specified paradigms, all participants obtained a classification accuracy that exceeded or was equal to 87.5%.

CONCLUSIONS

This study provides an optional route to enhance the event-related desynchronization activities and classification accuracy of a motor imagery brain-computer interface through optimization of motor imagery tasks and instructive guidance.