Key Technology Research On Resting State Functional Brain Network Modeling Based On Link Prediction

The human brain is one of the most complex system in the natural world, and it can be described by complex networks. At present, using complex network principles as an important means to research the brain network, researchers construct the brain networks under different scales and find the definition of the nodes and edges accord with the human brain mechanisms to draw the complete brain activity map.However, because of imperfect brain network constructing and analysis technology, there are still various urgent problems to be solved. For example,the work of the edges joint in the construction methods needs a lot of calculation, especially leads to huge costs consumption under the large scales node definition. Therefore, researchers need a reliable mathematical modeling methods to establish a generating function of brain network connection and continuously optimize in order to achieve the optimal fitting. The study not only excavates the relationship between the physical space distance and the cost and the network topology, but also has important significance to reduce the computational costs of network construction. In addition, the study leads to reliability problem of the brain network because of the various interference factors. The traditional analysis method based on the retest reliability can make consistency tests on network indicators, but doesn’t achieve the reasonable optimization of network structure itself. Therefore, there is an urgent need for reliability evaluation and the model optimization. The exploration research of this series issues provides a new thought and method for large scales brain network modeling and analysis. Meanwhile it has important theoretical value and practical significance.The current study includes the following research outcomes:(1)Modeling method of complex brain network based on local information index.The traditional local information index is introduces into functional brain network modeling with the brain anatomical distance to realize the mathematical modeling and simulation. And through the quantitative analysis to the network topology properties and reasonable model evaluation which are used to select the optimal model.(2)Modeling method of the complex brain network based on hierarchical structure model.The modeling based on hierarchical structure model with the characteristics of typical network modularity as the prior knowledge is used to complete brain network modeling. The results show that the modeling effect compared with traditional link prediction method is better on the time complexity.(3)The reliability optimization of complex brain network based on stochastic block modelIn previous studies, the core problem in the brain network is the quality control, namely the connection reliability evaluation. The traditional methods for quality control can only achieve the consistency measurement but lack the effective evaluation method for network connection. The current study convert the above problem to link prediction and proposes the brain network connectivity evaluation and reconstruction method. The results show that the reconstruction network is more close to the real brain network and improve reliability of the brain network.