Studies On Brain Network Based On Rhythms Of EEG Signals

The network analysis is a technique that concerns the connectivity across regions.It can be used in electroencephalogram (EEG) signal processing to reveal thephysiological mechanisms in terms of the inter-connectivity across different brainregions. In this dissertation, the method of network analysis other than the simple regionanalysis is used to study the brain rhythm signals. The network analysis based oncanonical correlation analysis (CCA) was proposed. In addition, functional connectivitypattern was also calculated using directed transfer function (DTF). A comprehensivestudy on both steady-state visual evoked potentials (SSVEP) and motor imagery (MI)was conducted in this study.The network analysis based on CCA was used in order to study the characteristicsof SSVEP. In the experiment, a novel visual stimulation paradigm was introduced. Theresults showedthat CCA network analysis could well discriminate the left and rightfrequency components which were projected to contralateral occipital regions. Based onthe results, an online brain-computer interface(BCI) system modulated by frequencyand space information was proposed. The attractive feature of this system is that it couldsubstantially increase the number of targets when using limited frequencies.The exact mechanism underlying SSVEP is not clear by now. The method basedon DTF was introduced to estimate the brain functional connectivity driven by SSVEPat different frequencies. The results indicated that the parietal region was a key neurallocus of human visual perception. Besides, we put forward the concept of flow gain forthe first time to explore the exchange and processing of brain information. Using flowgain mapping method, the functions for displaying and quantization of connectivitypattern were enhanced.The MI has always been a classical paradigm in BCI. In this study, the functionalconnectivity pattern during MI was also estimated. And detailed study were made to astroke patient. Compared with normal subjects, the results of patient showed that theconnections was stronger in the contralateral region. The damage of the motor regionmay be responsible for the results. Through a period of rehabilitation training, the results indicate that the stroke patient could using MI to control BCI system effectively.Moreover, we found that the functional connectivity pattern manifesting the dynamicbrain structure may serve as a reasonable method for rehabilitation assessment.Further studies on the estimated networks were conducted in this study. Themethod of network analysis in conjunction with graph theory would be helpful for theresearch on brain science.