Alzheimer's disease detection using sparse autoencoder, scale conjugate gradient and softmax output layer with fine tuning

Abstract

Accurate diagnosis of Alzheimer's disease (AD) plays an important role for patients care particularly in the early phase of the disease. Although numerous studies have used machine learning techniques for the computer aided diagnosis (CAD) of AD, an obstacle in the diagnostic efficiency was shown in the former methods, due to deficiency of effective strategies for characterizing neuroimaging biomarkers and limitation in choosing the learning models. In this study, we propose a deep learning model, which consists of sparse autoencoders, scale conjugate gradient (SCG), stacked autoencoder and a softmax output layer, to subdue the bottleneck and support the analysis of AD and healthy controls. Compared to the former workflows, out technique requires less labeled training examples and minimal prior knowledge. The proposed methods provides a significant improvement in classification output when compared to other studies, resulted in high and reproducible accuracy rates of 91.6% with a sensitivity of 98.09% and a specificity of 84.09%.