功能性磁共振成像（functional Magnetic Resonance Imaging, fMRI）技术目前已经被广泛用于考查人脑的认知神经加工机制。在各种fMRI数据分析方法中，静态功能网络连接（Functional Network Connectivity, FNC）可以揭示空间上分离但时域上耦合的脑网络间的交互关系。与静态功能网络连接分析相比，动态功能网络连接（Dynamic FNC, DFNC）分析方法可以捕捉随时间变化的脑网络动态属性，因此被越来越广泛的用于fMRI数据分析中。近年来，基于运动想象的实时fMRI神经反馈训练研究主要考察训练如何影响运动执行以及运动想象的神经机制。但是，运动加工的脑网络连接的动态属性如何随着实时神经反馈训练而变化尚不清楚。因此本文拟将采用DFNC方法探讨实时运动想象训练对运动加工的动态脑网络连接属性的影响。 研究一使用基于滑窗的动态功能网络连接分析方法，探究基于右侧背侧前运动区（premotor area，PMA）实时fMRI神经反馈训练如何改变手指敲击任务中脑网络连接的动态特性。研究发现，训练前后有两个重要的DFNC状态分别表征被试专注于手指敲击任务（状态6）以及被试注意力不集中的脑状态（状态2）。在训练后，实验组比控制组更多的保留在状态6，状态6的认知控制与默认网络间强度显著高于控制组；实验组较少的保留在状态2，状态2的默认网络强度显著低于控制组。结果表明真实的神经反馈训练使得实验组较少处于注意力不集中状态，更多的处于专注任务的状态，从而有效的提高了手指运动的行为表现。 研究二使用基于隐马尔科夫模型（Hidden Markov Model，HMM）的动态功能网络连接分析方法，探究基于右侧背侧PMA实时fMRI神经反馈训练如何改变手指运动的脑网络连接的动态特性。研究发现，神经反馈训练使得实控两组人在训练后分别更多地处于状态4以及状态7。在训练后实验组的状态2、状态4中分别发现，认知控制网络与默认网络之间和默认网络内部的子网络强度产生相同的变化趋势，体现出认知控制网络和默认网络是通过协同作用来协助被试完成实验任务。此外，经过神经反馈训练，实验组的状态4的持续时间以及运动网络内部的强度均和行为表现呈现显著正相关。结果表明神经反馈训练使得状态4成为实验组完成手指敲击任务中重要的脑网络连接状态，且该状态与行为之间呈现相关性。 两个研究共同揭示了基于右侧背侧前运动区的实时fMRI神经反馈训练能够改变被试所处功能连接状态的时间，且训练建立了功能连接状态的动态属性与行为之间的相关关系。

Functional magnetic resonance imaging (fMRI) technology has been widely used to investigate cognitive neuro mechanism of human brain. Among various fMRI data analysis methods, static functional network connectivity (FNC) analysis could successfully reflect interaction relationship between brain networks spatially separated but temporally coherent. Compared with static functional network connectivity analysis, dynamic functional network connectivity (DFNC) analysis method could capture dynamic properties of brain networks over time and was more and more widely applied to fMRI data analysis. In recent years, real time fMRI neurofeedback training based on motor imagery mainly investigated how the training influenced the neuro mechanism of the motor execution and motor imagery. However, it is still unclear how the dynamic properties of brain network connection in motor processing was changed by real time neurofeedback training. Therefore, this study aimed to investigate the effect of real time motor imagery training on dynamic brain network connection attributes in motor processing by DFNC analysis method. Firstly, we used DFNC analysis method based on sliding window to explore how real time fMRI neurofeedback training based on right dorsal premotor area change the dynamic properties of brain network connection during finger tapping task. The results showed that there were two important DFNC states represent the brain state in which subjects concentrated on the finger tapping task (state 6) and the brain state in which subjects were distracted (state 2) respectively. Compared with the control group, the experimental group spent more time on state 6, and the strength between cognitive control network (CCN) and default mode network (DMN) of state 6 in the experimental group was significantly higher than that in the control group. The experimental group spent less time on state 2 after training, and the DMN strength of state 2 in the experimental group was significantly lower than that in the control group. The results demonstrated that the real neurofeedback training made the experimental group stay less on the distracted state, but more on the state of the concentration task, which may effectively improve the behavioral performance of the finger movement. Secondly, we used DFNC analysis method based on Hidden Markov Model (HMM) to explore how real time fMRI neurofeedback training based on right dorsal premotor area change the dynamic properties of brain network connection during finger tapping task. The results showed that the neurofeedback training made the experimental group spend more time on state 4 and the control group spend more time on state 7 respectively after training. For the experimental group, the strength between CCN and DMN and the strength of DMN in state 2 and state 4 showed the same variation trend after training respectively, revealing that the cooperation between CCN and DMN contribute to the finger tapping task. Moreover, for the state 4 of the experimental group, the mean dwell time and the strength of motor subnetwork were significantly positively correlated with the behavior performance after the neurofeedback training. The results demonstrated that neurofeedback training made state 4 become an important brain network connection state in finger tapping task and the state 4 showed correlation with the behavior for the experimental group. Two parts of this study revealed that real time fMRI neurofeedback training based on right dorsal premotor area could change the time of functional connection states in which subjects stay, and training generated the correlation relationship between the dynamic properties of functional connection states and the behavior performance.