Research On Information Filtering Algorithm Of Depression Function Brain Network

The human brain,as the most sophisticated organ,requires about 20% of the total energy consumed by the body for its metabolism.The brain is internally composed of different regions and connected by neurons to form a complex network structure,which needs to find a balance between meeting functional demands and saving energy costs.While optimizing the topology to accommodate complex real-world activities,the organization of functional brain networks also needs to meet low wiring costs.As a result,brain network structures tend to exhibit a small-world nature,a structure that enables both local information processing and ensures global information integration.Research on the brain needs to focus not only on the activity of brain regions at rest or during specific cognitive activities,but also on the interactions between different brain regions.Functional Connectivity can reflect intercortical interactions without considering the causal relationships between cortical regions,and is one of the common methods to study intra-and inter-regional connections in the brain.Optimizing the fully connected functional connectivity matrix from the correlation analysis into a sparse network structure by binarization method is a key step in the subsequent network structure analysis.However,all existing binarization methods suffer from problems such as changing the original topology of the graph or assuming that the graph has specific constraints.To address these problems,this paper selects the group of patients with major depression as the research object,analyzes the MDD brain functional network connectivity pattern in depth,and applies it to the major depression dataset for experimental validation.The main contributions and innovations of this paper are as follows.(1)To solve the problem that existing binarization methods can hardly preserve the original structure and specific constraints of graphs,this study proposes a Trade-off Model between Cost and Topology under Role Division(MCT),which consists of three steps: community detection,the model consists of three steps: community detection,role division of nodes and edges,and E-cost dynamic optimization algorithm.This paper uses a synthetic dataset to evaluate the difference between the MCT method and other binarization methods in terms of denoising effect,and finds that the MCT method can better preserve the true connections present in the original data.The MCT method also showed better performance on an actual dataset consisting of 71 depressed patients and 51 healthy controls,suggesting that it preserves the underlying network structure while being more consistent with the small-world nature of functional brain networks.In addition,the MCT method was binarized on another real dataset to compare the differences between the two groups in terms of hub nodes,abnormal functional connectivity,and network characteristic parameters.The applicability and validity of the MCT method in the analysis of functional brain networks are further verified by comparing the results with those of related studies.(2)The MCT method has the problems of high computational cost and E-cost optimization algorithm which is not suitable for introducing the three role characteristics.In order to enhance the practicality and theoretical integrity of the MCT method,a trade-off model between Cost and Topology under Probabilistic Awakening(MPA)based on role division is proposed in this paper.The MPA model retains the MPA model uses probabilistic awakening instead of dynamic assignment,while retaining the ideas of community detection,role and edge division,and different types of connection assignment of the MCT approach.With the optimization of the probabilistic wakeup method,the running time of MPA is optimized to be close to that of the simple binarization method,while retaining the complex operations such as role delineation and edge wakeup,which have higher practical value.In addition,to verify the effectiveness of the MPA method,the network features generated by five different binarization algorithms are compared and analyzed in this paper.The results show that the results of the MPA method with one time the number of node scale wakeups are very close to those of the MCT method;while the MPA method with ten times the number of node scale wakeups performs better than the MCT method.After constructing the functional connectivity matrix for the major depression group and the healthy control group,this paper extracted three network feature parameters and used them as the input of three common classifiers,achieving a classification accuracy of91%,demonstrating the feasibility and effectiveness of the MPA method in real data.