The effects of cognitive parallel task with varying memory load on motor imagery BCI

Abstract

As one of the most practical EEG-based paradigms, motor imagery brain computer interface(MI-BCI) is not only used to control external devices, but also to help patients with hemiparesis to reconstruct impaired motor function. However, due to event-related (de)synchronization(ERD/ERS) during motor imagery are not stable enough, the performance of classification of MI-BCI is relatively poor. It has become a focus of study that how to achieve the feature enhancement of motor imagery. Since motor imagery is a cognitive processing that engages parts of the motor resources, we try to improve suppression and enhancement of amplitude (ERD/ERS) in a novel way by changing cognitive state. In this study, we designed a cognitive parallel n-back task with varying memory load to carry out with motor imagery task synchronously. The result of 13 subjects who volunteered in this experiment was shown that increased memory load could activate much stronger power decrease of ERD pattern at alpha rhythms in both sides of sensorimotor cortex, especially in contralateral area. Furthermore, we calculated the accuracy of classification between motor imaginary and motor idle status in different conditions by two classifiers, respectively. Through the paired t-test, we obtained that the accuracy of high memory load condition was significantly higher than the low load condition(SVM: (76.3±13.3)% and (83.4±10.5)%, p<0.01; LDA: (78.0±13.5)% and (84.6±12.4)%, p<0.05). A conclusion can be drawn that memory load have a positive impact on ERD pattern, even it is not caused by motor imagery itself. Besides, it may imply a new approach to modulate brain oscillations related to motor imagery by changing cognitive state.