Research On Connectome Construction Optimization And Local Features Visualization

This topic is part of the project performed by a research group of Indiana UniversitySchool of Medicine (IUSM group), which focus on solving some of the bottlenecksregarding the construction and analysis of large numbers of brain networks.The human brain is one of the most complex systems in the world which can beviewed as a network containing countless neurons as nodes and axons as edges. Owing tonon-invasive neuroimaging techniques, whole brain networks are constructed andmodernnetwork approaches are applied to explore the topology and dynamic principle of the brain,which has become an emerging research field, called Human Connectome.The aims of this thesis is to meet the needs of IUSM group about the HumanConnectome project, including:1)determining a network definition which can fully presentthe brain structure;2)accelerating the network construction algorithm to enable mass datacomputation;3)designing a visualization approach to facilitate intuitive analysis of largenumbers of complex connectome data;4) developing a volume rendering method tovisualize network features in the context of anindividualbrain anatomy.According to the requirements above, respectively, we:1) implemented a multi-scaleconnectome construction program(common effort with IUSM group), which generatesmulti-scale networks, presenting the whole brain in different networkresolutions;2)optimized the construction algorithm N(number of nodes,1015at most)times faster by using a lookup table approach to eliminate the computational redundancy;3)designed a brain network visualization scheme by extracting and rendering a series of localnetwork measures(e.g., nodal degree, strength and efficiency) from the connection matrixand providing a set of visualization strategies to facilitate the exploration and comparison of connectomes in different groups, scan-times and scales;4) proposed a new volumerendering methodto real-timehighlight individual network nodeswithin theoriginalbrainvolume by offline expanding the voxel intensity field to allowthe encoding of thenode indexes inside voxel values of the volume.The programs we built have been used to construct and analyze large numbers of realconnectome data from Indiana Universe School of Medicine. A paper related to this thesishas also been accepted by conference of EuroVis2012(First author).