Research On The Identification Method Of The Hub Nodes Of Brain Network Based On Graph Theory And Its Application In Neurological Diseases

Recent developments of brain imaging and graph theory allow us to investigate the complex functional and structural connectivity between brain regions and understand the pattern of how our brain work by using the tool of brain network.Mounting evidence has shown that brain network is characterized with the property of small-world and free-scale,which suggest that there only a small number of brain regions with highly connectivity(commonly called as hub node).Meanwhile,many studies have found that many of neurodegeneration were associated with the disruption of topological of brain network.Since the central role of hub node in brain network,it makes the hub nodes particularly susceptible to pathological network alterations.Thus,hub identification has attracted much attention in neuroimaging.Currently,many of hub identification methods have been proposed.Most of them selected hub node based on one or multiply graph centrality measures,which simply sequentially select a set of hub nodes by ranking a nodal centrality.However,most of these approaches are belong to a heuristic manner,which select hub node one-by-one independently without considering the synergies existed among a set of hub nodes.To overcome these limitations:(1)Aiming at the dynamic characteristics of functional brain networks,we propose a graph theory based dynamic detection algorithm for hub nodes of functional brain networks.We have evaluated our method on resting-state functional resonance imaging(f MRI)data from an obsessive-compulsive disease(OCD)study,where our new functional hub detection method outperforms current methods in terms of accuracy and consistency.(2)For structural brain networks,we propose a joint graph embedding and hub identification solution in a reinforcement learning framework to discover the unprecedented heuristics from the existing knowledge of network neuroscience and graph theory,which allows us to outperform the current state-of-the-art hub identification methods.We have achieved more reliable and replicable hub identification results on both simulated and real brain network data,suggesting the high applicability to various network analysis studies in neuroscience and neuroimaging fields.