| For many years,human beings have been devoting themselves to brain research.Among them,the construction of brain structure networks and the excavation of connection of brain networks have become the fields of natural sciences,brain sciences and neuroimaging research hot spots.In neuroimaging studies,understanding the pathology of brain diseases can be helped by exploring the interplay of structure and functional connectivities between brain regions.As brain network research gets more and more attention of scientists,it has become the research hotspot to construct the functional connectivity network and excavate the connection rules and topologies of brain network with different technologies.The resting state functional connectivity network has become one of the most popular technologies for building a human brain functional connectivity network.The resting state functional connectivity network can detect the nervous activity of the spontaneous low frequency of the brain,thus revealing the neural activity associated with brain disease.However,the conventional functional connectivity network computing the correlation based on the entire time series of RS-fMRI data simply measures the function connectivity between brain regions with a scalar value,which is fixed across time.This actually implicitly hypothesizes the stationary interaction patterns among brain regions.As a result,this method may overlook the complex and dynamic interaction patterns among brain regions,which are essentially time-varying and it is possible that the subtle complex and dynamic interaction patterns in the scanning time leads to a certain disease.High-order functional connectivity networks are rich in time information that can reflect dynamic changes in functional connectivity among brain regions.Accordingly,high-order functional connectivity networks have been widely used in the classification of brain diseases.However,traditional methods for processing high-order functional connectivity networks generally include the clustering method,which reduces the dimensionality of the data.As a result,such networks cannot be effectively interpreted from the perspective of neurology.Additionally,due to the large scale of high-order functional connectivity networks,it can be computationally very expensive to use complex network or graph theory to calculate certain topological properties.Here,we propose a novel method of generating a high-order minimum spanning tree functional connectivity network.This method increases the neurological significance of the high-order functional connectivity network,reduces the network computing consumption,and produces a network scale that is conducive to subsequent network analysis.After constructing the high-order minimum spanning tree functional connectivity networks,at present,the methods based on traditional local network features are widely used to analyze and classify brain networks.But this method has an obvious deficiency is some useful network topology information in the network(including connection patterns in the sample itself and the common connection patterns between the samples)would be lost,resulting in reduced classifier performance.The use of subgraph pattern as features just makes up for this defect.However,it is noteworthy that both methods based on the traditional quantifiable local network features or using subgraph patterns as features will have the loss of sample information.To ensure the quality of the topological information in the network structure,we used frequent subgraph mining technology to capture the discriminative subgraphs as features,and combined this with traditional quantifiable brain region features.Then,we applied a multi-kernel learning technique to the corresponding selected features to obtain the final classification results.Therefore,the performance of high-order minimum spanning tree network was verified from the perspective of local network features and discriminative subgraph patterns,which are captured by frequent subgraph mining. |
PDF Full Text Request