The study of brain neural system has a long history. A deeper understanding and a better utilizing of thinking is always our inner desire and motivation. With the advancement of related technologies, Brain-Computer Interface (BCI), a system aims to bridge the brain and the external environment, came into being and has been a hot interdisciplinary research field. Exploring the intrinsic relationship between the activity of neural system and the information it contains is still a basic and core issue in BCI studies. According to its inherent structure characteristics, it is meaningful to analyze the integrated function of neural system from the aspect of neural interaction networks. Therefore in this dissertation, our work was mainly based on BCI experimental platform, simultaneously recording spike train data when the subject was trained to complete multi-targets reach-to-grasp tasks. Different methods for evaluating neural interactions were investigated and neural network properties were further analyzed. The contributions of this dissertation are as follows: 1) According to reach-to-grasp tasks, we proposed a framework based on multivariate vector autoregressive model (MVAR) and partial directed coherence (PDC) for evaluating neural interactions, aiming to advance the independent analysis of single neuron to the neural networks level. Based on BCI experimental platform and implanted microelectrode arrays, spike train data was simultaneously recorded when a monkey completed multi-targets reach-to-grasp tasks. In the result, topology differences were observed for neural networks corresponding to different tasks. And generally existed neural interactions supported our previous hypothesis of common pathways in motor cortex for both transport and manipulation. 2) We proposed a Dynamic Bayesian Networks (DBN) model based method to evaluate neural interactions during reach-to-grasp, in order to better capture non-linear dependencies and dynamic properties between neural activities. The results provided further evidences for the collaborative operating mechanism of neural system. And the results also indicated that there might be some latent but not distinct rules for neural interaction networks to encode task information. In addition, the complexity differences were observed for neural interaction networks according to delay period and peri-movement period. And in the investigation of within-trial change rules of neural interaction networks, different degree...
修改评论