Optimizing The Reliability Of Functional Brain Network In Resting State Based On Stochastic Block Model

The human brain is the most complex and mysterious organization in the currently explored universe. The Advanced Function of brain is the most complex Motion style in nature. Its complexity is reflected in building in different scales and connections reflect different modes in Cognitive function,thoughts, feelings and behavior. So the working principle of exploration and research on the brain and logic has become the contemporary natural sciences one of the biggest challenges. In recent years, more and more scientific workers pay attention to and aware of the necessity and importance of the construction of the brain network.The study surrounds the international hot area of research of the brain network closely, and in-depth study of resting state functional brain network build and reliability of brain networks. In this paper, the complex network research adopted the idea of study, through the link prediction algorithm to optimize the reliability of brain networks.The main innovation works are as follows:(1) Evaluating the performance of link predictors of brain networks.According to link prediction, edge set of brain networks which the first part of the data construct were divided into training set and testing set. Using link prediction evaluation index AUC and precision to compare indicators’ performance of the link prediction,such as CN、PA、RA and SBM. Results show that the performance of the stochastic block model is better than others.(2) Verifying the applicability of reconstruct on brain network basing on the stochastic block model. Brain networks which first part of the data construct are considered to be real network that there is no error edges in network,randomly deleting and adding edges of real networks to form observed networks,and then using the stochastic block model to reconstruct observed networks to form reconstructed networks. The relative error of attribute between observation network and real network, and the relative error of attribute between reconstruction network and real network are Computed. Comparing the relative error of attribute show that the network reconstruction based on random block model is suitable for the resting state cerebral network.(3) Optimizing the brain network structure used the stochastic block model.Reconstructing brain networks which second part of the data construct.Computing retest reliability of attributes of observation and reconstruction networks. Comparing retest reliability of attributes of observation and reconstruction networks to measure the performance of optimize brain network by reconstructing network. The results show that the retest reliability of brain network improves to some extent that means improve the reliability of the brainnetwork, when reconstructing the retest reliability lower brain networks.The study researches the reliability of brain networks using the thought of link prediction, and found more suit index of link prediction to predict brain networks by comparing common indexes of link prediction. Network reconstruction in brain network is verified Basing on stochastic block model.Lower reliability of brain networks are reconstructed basing on stochastic block model, the results of the reconstruction is verified by using test-retest reliability.And results show that reconstruction can improve the reliability of brain network. Network reconstruction also have some reference values for Exploring the way brain network data measurement and calculation method of brain network building.