Construction And Characrsitc Identification Of Brain Network In Alzheimer’s Disease

Electroencephalograph and magnetic resonance imaging are effective tools for clinical diagnosis of Alzheimer’s disease.However,the identification accuracy is usually not satisfactory due to lacking effective neuroimaging characristics.Neuroimaging data are quantified with the combination of modern information processing and machine learning methods for assessment of brain function in patients.It can provide a valid basis for the diagnosis of Alzheimer’s disease.Electrophysiological experiments are designed to acquire EEG data of Alzheimer’s disease patients.Permutation disalignment index is used to estimate the functional connectivity between each pair-wise EEG signals and to construct functional brain networks.Complex network theory is further applied to analyse the alteration of brain network structure and extract the features.Comparing with healthy controls,correlation between the brain regions is weakened and the small world index of the functional network is reduced in Alzheimer’s disease.It indicates a decline in the ability of infomation transmission and processing between brain areas.A novel machine learning method network-based Takagi-Sugeno-Kang(N-TSK)is proposed with the combination of function brain network characteristics and fuzzy classification algorithms.The topological characteristics of weighted and unweighted networks are extracted.Taken the network parameters as independent inputs,N-TSK model is established and further trained.The results demonstrate the effectiveness of the proposed scheme in Alzheimer’s disease.The performance of weighted network which the highest accuracy can achieve 97.30% is largely exceeds unweighted network.Furthermore,it is found that local efficiency and clustering coefficient are most effective factors in Alzheimer’s disease identification.With the spatial structural characteristics of magnetic resonance imaging,the brain is divided into network nodes,and the structural similarity between nodes is analyzed to reconstruct the brain network of Alzheimer’s disease patients.TSK,k-nearest neighbor,support vector machine and other machine learning algorithms are used.The results show that TSK had the best performance in identifying the brain network characteristics in Alzheimer’s disease with the accuracy 94.55%.The node betweenness and the edge betweenness are the most effective characteristic parameters,which indicate that the brain spatial structure of the Alzheimer’s disease is affected.This paper proposes a brain network-based identification method for Alzheimer’s disease.The brain networks are constructed based on neuroimaging data.Then the Alzheimer’s desease is identified by using machine learning approaches.This method provides a basis for brain function assessment and clinical diagnosis of Alzheimer’s disease.