Research On Feature Selection And Classification Methods For Alzheimer's Disease

In recent years,the number of elderly people is increasing,and the incidence of geriatric diseases has increased significantly.The most representative of diseases is Alzheimer's Disease(AD),which is often referred to as Alzheimer's disease.Data display that the average survival time of Alzheimer's disease patients is only 5.5 years,which is the “fourth killer” that affects the health of the elderly after cardiovascular disease,cerebrovascular disease and cancer.According to conservative estimates by the International Federation of Alzheimer's Disease,the number of global Alzheimer's patients will increase to 75.62 million by 2030.By 2050,the number of patients will reach 135.46 million.It is urgent to early diagnosis and prevention of Alzheimer's disease.Mild Cognitive Impairment(MCI)is generally considered as a transitional state between normal control(NC)and Alzheimer's disease.Some studies have further classified MCI patients into cMCI(converted MCI)and sMCI(stable MCI)according to whether MCI is converted to AD within a certain period of time.Statistics show that cMCI patients are more likely to convert to AD patients within a certain period of time.So the study of MCI patients in this article is mainly cMCI patients.Although cMCI patients are relatively normal in the initial clinical stage,they will eventually be converted to AD patients if they are not treated in advance.If the population of AD,cMCI and NC can be accurately identified in the crowd,it will prevent and interfere with AD disease to a certain extent.The main research content of this article includes:(1)The SVM-RFE feature selection method is based on the weight vector generated by the SVM during training to construct the sorting coefficient,and removes the smallest feature of the sorting coefficient at each iteration to implement feature selection.This method considers only the feature and the target class.The correlation does not take into account the redundancy between features.In order to make the SVM-RFE feature selection method consider the redundancy between features,the relevant algorithm in mRMR is introduced to reconstruct the sorting coefficient after the SVM-RFE feature selection method is used to generate the weight vector W.It completes the improvement of the SVM-RFE feature selection method and forms a new feature selection method.(2)In order to verify the effectiveness of the improved SVM-RFE feature selection method,the paper used the SVM classification method and the KNN classification method to classify cMCI and NC population,AD and NC population based on cortical thickness data obtained from a series of MRI brain imaging data of AD,cMCI and NC.The results show that the accuracy of the classification using the improved SVM-RFE feature selection method is not only higher than the accuracy of the three methods using the SVM-RFE feature selection method,the mRMR feature selection method,and the F-score feature selection method,but also the number of feature subsets is also the least of the three methods.