Alzheimer's disease (AD) is one of the most influential nervous system diseases in the world. It is accompanied by symptoms such as loss of memory, thinking, and language ability. This paper discusses the characteristic indexes of brain magnetic resonance imaging (MRI) in mild cognitive impairment (MCI) and AD. It applies the MRI characteristic indexes in machine learning to classify and predict the course of AD to select the best model for classification and prediction auxiliary diagnosis of AD. In this study, 560 eligible subjects numbered 0-15,000 in the AD Neuroimaging Initiative (ADNI) database were randomly selected. According to the ADNI diagnostic criteria, the subjects were divided into four groups: the cognitive normal (CN) group (n=140), 230 cases in the early MCI (EMCI) group, 110 cases in the late MCI (LMCI) group, and 80 patients in the AD group. Random forest (RF), decision tree (DT), support vector machine (SVM) algorithms were used to classify and predict the different disease progress of AD. Next, different MRI indexes were input into the three machine learning algorithms to predict CN-EMCI-LMCI-AD. We compared the prediction accuracy, sensitivity, specificity, and area under the receiver operating characteristic (ROC) curve (AUC). This study found that CN-AD had the highest classification accuracy, followed by EMCI-AD, CN-LMCI, LMCI-AD, EMCI-LMCI, and CN-EMCI. In the prediction of CN-AD, the AUC of 0.92 of the RF classifier was higher than the AUCs of the SVM and DT classifiers. Of the three machine learning algorithms, RF was better than the SVM and DT at predicting different MRI features. The accuracy of RF, SVM, and DT was 73.8%, 60.7%, and 59.5%, respectively. The RF classifier had the best prediction effect on different disease processes of AD. Five MRI indexes (used as classification features) had the best prediction effects. CN-AD had the best classification effect. Overall, the classification accuracy of the RF classifier for CN-EMCI-LMCI-AD was higher than those of the other models. The RF classifier can be used to classify different stages of AD in the early stages of the disease to assist in diagnosing AD.
Read full abstract