Classification Of Alzheimer's Disease Based On Multimodal MRI Features Analysis

Alzheimer's disease(AD)is one of the most common neuropathic disease that occurs in middle aged people.There is neither quantitative diagnostic method nor treatment method that can cure the disease completely.As the aging problem worsens,the diagnosis and treatment of the disease need to be further improved.Magnetic Resonance Imaging(MRI)can quantitatively reflect the change of brain tissue on the structure and function,and have the characteristics of noninvasiveness,high spatial resolution.MRI are widely used in the diagnosis of AD.With the gradual maturity of multiple MRI sequences,the application of multi-characteristic MRI technology plays an increasingly important role in the diagnosis of AD.This paper proposes a pattern recognition research method based on brain structure information and multi-feature fusion,and explores its application in AD diagnosis.The main work of this paper is as follows:Build a classifier model.A classifier model was built.Based on Support Vector Machine(SVM),a basic classifier model with packaged feature selection and parameters optimization was built to be used in the optimization of the AD diagnosis,using the recursive feature elimination algorithm(SVM-RFE)for feature selection,the optimal feature subset selection,and reduction of the feature dimension.Using the gradient filtering algorithm to optimize the parameters of SVM,the performance of the classifier was improved..Brain network feature fusion method is applied in AD diagnosis.Build good classifier model,through the network connection structure characteristics,analysis of gray matter abnormalities,the result confirmed that is superior to the direct method of analyzing the characteristic of the gray matter,more help to AD disease early detection research;By comparing the analysis results of nodes and links,confirmed that the spatial distribution of the loss of connection in the brain structure in patients with AD and gray matter changes in the structure of inner link,further shows the based on the structure of the connection mode analysis in the diagnosis of early pathological and potential.Multimodality MRI data fusion is applied in AD diagnosis.Compared with the accuracy of AD classification in DTI,DKI and fMRI,the classification accuracy of DKI was up to 91.84% and the accuracy of fMRI was 73.47%.After the fusion,except the fMRI mode,the accuracy of the other two modes did not increase.The ROI information of the participant was explored,and the brain region associated with AD was found.In terms of structure,the core area was the area of the hippocampus and the brain island.The core area of functional area is the frontal gyrus and temporal gyrus.This paper constructs an optimized classifier model,studies the brain structure feature fusion method,and explores the application of multiple modality fusion in AD diagnosis.It is proved that the fusion of multiple features can improve the diagnosis method of AD,and provide some theoretical basis and method basis for relevant clinical or practical application.