Research On The Dimension Reduction And Visualization Platform Of Brain Network State Observation Matrix Based On T-SNE Algorithm

Functional magnetic resonance imaging is an effective method for the study of function and structure properties of human brain.The brain functional network reconstruction technology with blood oxygen level dependent signal based on functional magnetic resonance imaging provides an effective method for researching the dynamic features and characteristics of human brain.In order to study the dynamic characteristics of brain network,the brain area correlation analysis was carried out with the time series of blood oxygen level dependent signals and the whole brain network state observation matrix was constructed.Because the matrix is of high dimension,it is very difficult to identify the main dynamic properties of human brain network.Therefore,the effective dimensionality reduction method of brain network state observation matrix is the basis of further study for dynamic characteristics of brain network.Taking the brain network state observation matrix as the object of our research,we studied how to realize it's embedding from the high dimensional space to the low dimensional space and visualization in low dimensional space.Considering the embedding results with those dimensionality reduction algorithms such as principal component analysis,isometric mapping and local linear embedding are not applicable for resolving our problem.The characteristics of the data are studied and analyzed deeply,then a dimensionality reduction method for the high dimensional brain network state observation matrix based on t-distributed Stochastic Neighbor Embedding algorithm is presented in this thesis.The experiments results show that compared with other dimensionality reduction algorithms,this method can make a more effective dimensionality reduction and give a better visualization in low dimensional space for the brain network state observation matrix.On the basis of this method,we designed and implemented a dimension reduction algorithm and visualization platform with Python,and combined this platform with existing brain network information processing platform integration.All these works provide the theoretical basis and technical basis for the further research of the dynamic characteristics of human brain network.