Connectivity Analysis of Brain States and Applications in Brain-Computer Interfaces

Human brain is organized by a large number of functionally correlated but spatially distributed cortical neurons. Cognitive processes are usually associated with dynamic interactions among multiple brain regions. Therefore, the understanding of brain functions requires the inves- tigation of the brain interaction patterns. This thesis contains two main aspects. The first aspect focuses on the neural basis for cognitive processes through the use of brain connectivity analysis. The second part targets on assessing brain connectivity patterns in realistic scenarios, e.g., in-car BCI and stroke patients. In the first part, we explored the neural correlates of error-related brain activity. We recorded scalp electroencephalogram (EEG) from 15 healthy subjects while monitoring the movement of a cursor on a computer screen, yielding particular brain connectivity patterns after monitoring external errors. This supports the presence of common role of medial frontal cortex in coordinating cross-regional activity during brain error processes, independent of their causes, either self-generated or external events. This part also included the investigation of the connectivity during left/right hand motor imagery, including 9 healthy subjects, which demonstrated particular intrahemispheric and interhemispheric information flows in two motor imagery tasks, i.e., the ÎŒ rhythm is highly modulated in intrahemispheric, whereas β and γ are modulated in interhemispheric interactions. This part also explored the neural correlates of reaction time during driving. An experiment with 15 healthy subjects in car simulator was designed, in which they needed to perform lane change to avoid collision with obstacles. Significant neural modulations were found in ERP (event-related potential), PSD (power spectral density), and frontoparietal network, which seems to reflect the underlying information transfer from sensory representation in the parietal cortex to behavioral adjusting in the frontal cortex. In the second part, we first explored the feasibility of using BCI as driving assistant system, in which visual stimuli were presented to evoke error/correct related potentials, and were classified to infer driver’s preferred turning direction. The system was validated in a car simulator with 22 subjects, and 7 joined online tests. The system was also tested in real car, yielding similar brain patterns and comparable classification accuracy. The second part also carried out the brain connectivity analysis in stroke patients.We performed exploratory study to correlate the recovery effects of BCI therapy, through the quantification of connectivity between healthy and lesioned hemispheres. The results indicate the benefits of BCI therapy for stroke patients, i.e., brain connectivity are more similar as healthy patterns, increased (decreased) flow from the damaged (undamaged) to the undamaged (damaged) cortex. Briefly, this thesis presents exploratory studies of brain connectivity analysis, investigating the neural basis of cognitive processes, and its contributions in the decoding phase. In particular, such analysis is not limited to laboratory researches, but also extended to clinical trials and driving scenarios, further supporting the findings observed in the ideal condition.