Recognition of Alzheimer's Disease Using Structural MRI Based on Smooth Group L1/2

Abstract

Accurate classification of Alzheimer's disease (AD) is helpful for timely taking relevant measures in the early stage of AD, controlling the incidence rate of AD in key population and delaying the deterioration of AD disease. In this study, the calibration support vector machine (c-SVM) model based on smooth group L1/2 (SGL1/2) was used to select the key features of key brain regions, so as to realize the prediction and auxiliary diagnosis of AD. In the experiment, this method is applied to structured magnetic resonance imaging (s-MRI) datasets for training and testing. Compared with other group level regularization methods, the classification model of SGL1/2 combined with c-SVM has better effect on AD recognition. The conclusion of this study provides an objective reference for the automatic diagnosis of AD in the future.