The Application Of Tensor Decomposition Model In The Diagnosis Of Brain Magnetic Resonance Imaging

In recent decades,the rapidly developing science and technology are making it much easier to collect and store data.And along with the arrival of the era of big data,people are exposed to a great of data from different fields every day,including tensor.With the growing demand of quality and speed in data processing,more and more researchers are paying attention to the tensor data,and start to explore the operation rules and decomposition algorithms of it.Meanwhile,as a kind of high-order tensor data,the magnetic resonance imaging(MRI)has been widely used in computer aided diagnosis because of its low radiation and high resolution.Therefore,how to combine the tensor decomposition model with magnetic resonance imaging to accomplish computer aided diagnosis for certain diseases,has increasingly become one of the hot spots at home and abroad.The main research content of this paper is the application of tensor decomposition model in brain magnetic resonance imaging diagnosis.Starting with the basic concept and operation rules,this paper applies the existing tensor decomposition algorithms to the classification of magnetic resonance images and disease diagnosis,and then put forward our new tensor decomposition algorithm.We primarily introduce the concept and research status of tensor decomposition algorithm,and then present the frequently-used magnetic resonance sequences and a complete set of preprocessing procedure.Next we research the auxiliary diagnosis of Alzheimer's disease and put forward our new tensor decomposition algorithm.At last,the classification and diagnosis of glioma and related diseases are discussed.The main contributions and innovation of this paper are in the following aspects:Firstly,sum up the basic concepts and operation rules of tensor,and then introduce some classical tensor decomposition algorithms such as CP and Tucker.Meanwhile,generalize the basic principles and features of magnetic resonance imaging,summarize the complete set of preprocessing procedure,and build the data set of glioma and related diseases.Secondly,put forward a new tensor decomposition algorithm—Tensor Optimal Scoring(TOS),and use it in the diagnosis of Alzheimer's disease.We start with classical Linear Discriminant Analysis(LDA),analyze its small-sample problem,lead to the concept of Optimal Scoring,and then extend its scope of use to tensor data,put forward Tensor Optimal Scoring,overcoming problems such as the loss of effective information and the "curse of dimensionality".In order to verify the effectiveness of our new method,we use the OASIS dataset of Alzheimer's disease to design control experiments.Compared with other algorithms,our algorithm provides better results.Thirdly,propose a new diagnosis method of glioma based on Multilinear Principal Component Analysis(MPCA).We summarize the preprocessing procedure of the magnetic resonance images of glioma and related diseases,and then analyze the basic principles and operation steps of MPCA,use it in the diagnosis and classification of glioma disease.Besides,in order to make full use of effective data and promote the accuracy of algorithm,we fuse different series of magnetic resonance images and extract features,taking advantage of MPCA in dealing with high-order tensor data.Then we validate the method in the established glioma data set,the ideal results demonstrate the direction of future research.