The Research Of Functional Magnetic Resonance Imaging Based On The Random Neural Network Cluster

The human brain is the command center of all activities,thus it is meaningful to understand the structure and function of brain.Resting state functional magnetic resonance imaging technology has been widely applied to study the structure and function of brain and provides more theoretical basis to treat the brain disease.The neural network is an effective tool to classify the model belonging to the machine learning which evolves from the human brain and can achieve very strong nonlinear mapping function.In this paper,we use the random neural network cluster method respectively to study Alzheimer's disease and Autism spectrum disorders based on the resting state functional magnetic resonance imaging.The main content is as follows:(1)The random neural network cluster is proposed.The single neural network has the advantages of small computation and easy to realize,but its classification performance exists the unstable problem.We randomly select samples and features,then the samples and features enter into the large number of neural networks which combine with forming a random neural network cluster.The random neural network cluster effectively solves the problem of information loss in the traditional dimensionality reduction,and the performance has been improved.(2)The functional connectivity of the brain is analyzed based on the random neural network cluster to study the Alzheimer's disease patients.Firstly,the Pearson correlation coefficient of the brain regards as the features of each participant.Then the five different neural networks are chosen which construct five random neural network clusters to classify Alzheimer's disease patients and the healthy control.Next,the classification accuracy of five random neural network clusters are compared,then the random Elman neural network cluster is chosen as the best base classifier and the classification accuracy is 92.3%.Finally,the abnormal functional connectivity and brain regions are found in Alzheimer's disease patients,thus it provides a new perspective for the diagnosis of Alzheimer's disease.(3)The functional connectivity of the brain is analyzed based on the random neural network cluster to study the Autism spectrum disorders patients.Firstly,the degree,aggregation coefficient,the local efficiency and the shortest path are integrated together based on the graph theory method which regard as the features of the brain.Then the five different neural networks are chosen which construct five random neural network clusters to classify the Autism spectrum disorders patients and typical controls.Next,the classification accuracy of five random neural network clusters are compared,then the random Elman neural network cluster is chosen as the best base classifier and the classification accuracy is 95%.Finally,the abnormal functional connectivity and brain regions which include the supplementary motor area,the median cingulate and paracingulate gyri,the fusiform gyrus and the insula are found in Autism patients therefore,it provides a new way for the diagnosis of the Autism spectrum disorders.