Classification Of Mild Cognitive Impairment Based On Resting State FMRI

Alzheimer’s disease(Alzheimer’s disease,AD)is an irreversible and chronic neurodegenerative disease with the clinical characteristics of cognitive function from slow to rapid decline.Mild Cognitive Impairment is an intermediate state between normal aging and dementia,and 10%-15% MCI patients are converted to AD each year.Currently clinical diagnostic criteria is lacking in this impairment.With the development of neuroimaging technology,more and more people are exploring the early diagnosis of MCI disease by constructing structural and functional brain network.In order to better understand the effects of different degrees of disease on the changes of brain function network and whether the brain function network is helpful for early diagnosis and classification of diseases,the main content of this paper includes the following two parts:(1)We constructed brain function network under different frequency bands to analyze the difference of brain functional network between different degrees of disease.First,the resting state functional magnetic resonance data of the normal control group(Normal Control,NC),the early MCI(early MCI,EMCI)group,the late MCI(late MCI,LMCI)group and the AD group were selected.Secondly,four groups of rs-fMRI data were preprocessed,and the filtering stage was divided into three frequency bands(fullband,slow-4 and slow-5).Then time series of the needed brain area were extracted and the brain function network was constructed by Pearson correlation coefficient(Pearson ’s correlation).Finally,we used graph theory to calculate and analyze brain network functional differences between EMCI and LMCI groups,as well as between AD and NC groups.The results show that there are no significant differences in global network properties between EMCI and LMCI and between AD and NC in the slow-4 band.In the full-band band,the global efficiency of EMCI and NC is significantly greater than LMCI and AD in a small part of the threshold,while the characteristic path length of LMCI and AD is significantly greater than that of the small part of the threshold.In the slow-5 band,the global efficiency,the local efficiency and the average clustering coefficient of EMCI and NC are significantly greater than those of LMCI and AD,respectively.Similarly,The LMCI and AD characteristic path length is significantly greater than EMCI and NC under most threshold values.(2)Based on the characteristics of functional network,we classified EMCI and LMCI,AD and NC two groups.On the EMCI and LMCI groups,we used mRMR,SSLR and FS three algorithms to select the features of network attributes,and used SVM classifier to implement classification.The classification results show that the features selected by the mRMR algorithm have higher classification performance than those selected by the SS-LR and FS algorithms.The classification results obtained by using mRMR algorithm in low frequency slow-5 with mRMR algorithm are the best,and the classification accuracy is 83.87%.In group AD and NC,using SVM and Decision tree based stepwise backward deleting feature selection algorithm,the method of feature selection and classification based on decision tree is better.The method of feature selection and classification based on SVM is less effective.Under different frequency bands,the classification results obtained by slow-5 at low frequency are the best,and the accuracy of classification is 94% by using decision tree.It can be seen that the classification results of different frequency bands are different,and the classification effect of slow-5 band is more stable than that of slow-4 and full-band bands.