Abstract:Prediction of Alzheimer's disease (AD) from Mild Cognitive Impairment (MCI) has become popular in recent years. Especially, deep learning technique has been used to extract high-quality features and for classification in this topic. Whether the patient would converse from MCI into AD is a particular evaluation criteria in clinics. However, there is no such a conversion prediction model in literature. Therefore, the purpose of this study is to propose a decision supporting model based on deep learning and machine learning to predict the conversion probability from MCI into AD within one year. We analyzed 165 samples with MRI scans from Alzheimer's Disease Neuroimaging Initiative (ADNI) database, in which all MCI patients were converted into AD in different time span for conversion. In this model, we first extracted image features based on convolutional neural network (CNN) method, and then we used support vector machine (SVM) classifier to classify these features. The results showed that the classification accuracy using linear, polynomial and RBF kernel could achieve 91.0%, 90.0% and 92.3%. As a result, this study indicated that the decision supporting model is potential to be applied into predicting the conversion probability from MCI into AD within one year.

Decision Supporting Model for One-year Conversion Probability from MCI to AD using CNN and SVM.