Research On Visual Analysis Method Of Weighted Brain Network

The study of human brain connectomics has been an important subject in the field of biology in recent ten or twenty years.Advances in medical imaging technology have allowed scientists to study the functional networks and organization of the brain in a noninvasive manner.As an important means of data analysis,visual comparison of human brain networks has gradually become an important task of brain connectomics.This paper discusses and studies the visualization method and the construction of brain functional network.Firstly,the current popular visualization methods for brain functional network are discussed and compared,and the advantages and disadvantages of each method are analyzed,and a new visualization method combining with static brain functional network is proposed.Second in brain function network build modular partition algorithm is analyzed,combined with the characteristics of the brain function network was improved,introducing Louvain algorithm with module degree constraint to division of brain function network,combined with the adjacency matrix and node link graph optimization visual representation form of brain function network,finally built a network of brain function visual analysis system verify the effectiveness of the method and application value.The main contributions of this study are as follows:(1)To solve the problem of single form of visual representation of brain functional network and low efficiency of obtaining useful information.Based on the Nodetrix visualization method combining adjacency matrix and node link,a superimposed visualization representation for brain network connection weighted graph comparison is proposed.This method solves the problem that the connection strength information and the brain region location information cannot be obtained simultaneously in the representation of the adjacency matrix and node link graph,and the problem that the left and right juxtaposition view has low accuracy and long time when doing the comparison task.(2)In the task of modular partition of brain functional network,Louvain algorithm excessively pursues the maximization of modularity,and thus ignores the problem of subfunctional areas in the brain functional network.A new Louvain algorithm with modularity constraint is proposed,which is more suitable for the modularization of brain functional networks.The iterative ending condition of the original Louvain algorithm is improved from the original maximum of modularity gain to the comprehensive consideration of all modularity distribution in the dynamic brain functional network.It solves the problem that the module degree of partition results is not reasonable.(3)To solve the problem of single function and low efficiency of the visualization tools currently used with brain functional network.A more functional visual analysis system for comparing brain functional networks was established,which could highlight the differences between groups.This system supports both two-dimensional and three-dimensional modes,and shows the differences of brain networks between groups in a more intuitive and visual way.Meanwhile,secondary clustering is conducted for brain regions between different groups,and the difference analysis of clustering results between normal people and patients is carried out to visually find the brain regions with the greatest changes.Finally,functional magnetic resonance imaging(f MRI)data from a group of Alzheimer's patients and healthy controls were used for case studies.A group of controlled experiments were conducted to verify the usefulness of Nodetrix superposition representation in the comparison task of brain functional networks,and to verify the rationality of the modular partition between the Louvain algorithm with module degree constraint and the final results of the original Louvain algorithm,and to verify the application value of the visual analysis system built in this study.