Investigation And Application Of Individual Eeg Power Spectrum And Brain Network Analysis Methods

There are huge individual differences in human brain neural oscillations and functional networks.Accurate acquisition of EEG rhythm and functional information are the prerequisite to study the neurobiological mechanisms of the individual brain with electroencephalography(EEG)data.Traditional EEG power spectrum analyses depend on predefined frequency bands(e.g.theta,delta,alpha,beta and gamma bands),which ignore the variability in these frequency bands across participants,reducing the effectiveness of the results.Besides,brain network has been considered as a static system and comparison between different groups using a constant value to characterize the average-connection of a certain period or local brain area,which ignore the individual differences and variability in different dimensions underlie the function network.Thus,emerging needs has been made to identify individualized EEG rhythms and abnormal connectivity patterns in multiple dimensions.Individualized analysis can also promote the clinical diagnosis and treatment,and become the determinant of precision medicine.In the present dissertation,we propose an individualized power spectrum algorithm and a dynamic analysis of individualized brain network.Then taking autistic and typical developmental children as the main subjects,we focus on the characteristics of individualized alpha rhythm and abnormal connectivity patterns in terms of spatiotemporal-spectral dimensions in brain network.Exploration and analysis of individual EEG data promote the research of neural mechanisms and EEG markers of disease.Furthermore,this investigation provide new ideas and methods for the study of neurodevelopmental disorders such as autism,and promote scientific guidance for early diagnosis and personalized treatment plan of autistic children.The work of this dissertation mainly includes the following three parts.(1)Automatic segmentation method of individual power spectrum bandsTo investigate the difference between individual spectrum power and distribution of frequency bands,we propose a novel individualized automatic segmentation of power spectrum method based on consensus clustering learning.The analysis of simulation and experimental data have confirmed the effectiveness of the method in extracting individual frequency components.The results demonstrate that the proposed method is promising for extracting individualized EEG frequency bands.(2)Individual network dynamic analysis methodsTo investigate the dynamic characteristics of the brain network in multiple dimensions,we established three progressive brain network dynamic models.For the temporal domain,a network microstate detection method is proposed to study the dynamic characteristics of brain network from the time scale.For the spatio-temporal domain,a stability-driven nonnegative matrix factorization(staNMF)method decomposes the network into a set of network space components with a specific terrain topology and a set of network time components in individual level.For the spatio-temporal-spectral dynamic characteristics of network,the high-dimensional network algorithm is based on tensor decomposition method,which decomposes the network into three parts,including a set of network space components with a specific terrain topology,frequency components in regulation each network and a set of network time components in individual level.Through the study of experimental and / or simulation data,the effectiveness of the above methods have been verified.(3)Application of individual EEG analysis in children with autismAutism spectrum disorder(ASD)is a common neurodevelopmental disorder in children,whose brain abnormalities are more manifested in individual differences in brain rhythm and abnormal connection patterns of brain networks.In regard to the brain rhythm and network dynamic characteristics in multiple dimensions,we focus on the exploration and investigation on the individualized alpha rhythm and network connection patterns of autistic children in individual level and various aspects with resting EEG data.The study of individualized alpha rhythm found that the range of individual alpha frequency(IAF)is 7-10Hz(traditional alpha band: 8-13Hz),and individual-level alpha rhythm characteristics can steadily distinguish autistic and typical developmental children.These results reveal the importance of individualized power spectrum analysis.The study of dynamic individualized network analysis found that the main connection patterns in autistic children are long-range/ inter-hemispheric under-connectivity and short-range/intra-hemispheric over-connectivity.The patterns of long-range connectivity are modulation by low frequency rhythm,while short-range are higher frequency rhythm.These results reveal that accurate acquisition of brain rhythm features and network connection patterns make great help to understand the neural mechanism of autistic children.Collectively,our individualized EEG analysis could offer specific potential electrophysiological markers,and provide personalized treatment plan for clinical regulation and intervention of autism.