Research On Brain Dynamic States Based On Neuroimaging

The brain is the most advanced part of the nervous system,and dominates all cognitive functions,such as language,movement,thinking,memory,perception and emotion.The brain exhibits various functional states in different time scales and periods,and functional changes occur all the time.Recently,revealing and studying the dynamcs of brain states have been of great interest to researchers in the fields of brain science and neuroscience,and relevant work is rather insufficient.Based on the advanced neuroimaging technology,brain dynamic states were revealed in this dessertation from mouse to human,and from freely behaving condition to task condition.The main contents of the dissertation have been put forward as follows:Given the fact that recent local field potential(LFP)acquisition methods cover only a few brain structural areas and have only a few channels,the dissertation proposed a multi-site,multi-channel and full-depth LFP acquisition method,by which super-long LFP signals with super-high temporal resolution were successfully collected from 13 classic structural regions of mouse brain.Based on the LFP signal acquisition method proposed in this dissertation,considering the complex and composite structure of functional connectome,the dissertation proposed a research method of brain dynamic state based on sparse coding method and atomic functional connectivity.With LFP signals,6 and 4 atomic functional connectivities were extracted respectively from alpha and theta frequency bands of freely behaving mice,and their dynamic transition characteristics were revealed.Based on the study of the brain dynamic state of freely behaving mice described in this dissertation,given the complex and changeable cognitive function states of brain,the dessertation proposed a recognition method of brain dynamic state.Based on LFP signals,with the methods of Bayesian connectivity change point model and recurrent neural network,the dynamic characteristics of the four states of mouse brain in response of stimulation were successfully revealed.Based on the studies of mouse brain dynamic state described in this dissertation,considering the complex structure and the rich and changeable cognitive activities of human brain,the dessertation proposed a human brain dynamic state recognition and analysis method.Based on task functional magnetic resonance imaging data,the brain dynamic states of 784 participants under 7 tasks were identified accurately by a novel five-layer deep sparse recurrent neural network model,with an average accuracy of over 90%.