Alzheimer's disease diagnosis from structural MRI using Siamese convolutional neural network

Deep learning (DL) methods have been recently utilized in medical imaging diagnosis and prognosis, which have significantly improved the performance of algorithms. As Alzheimer's Disease (AD) is one of the most financial costly diseases, many researchers have concentrated on introducing a high accuracy automated algorithm for classifying the AD and the Normal Control (NC) cases. In this paper we proposed a new deep learning based automated method for Alzheimer's disease diagnosis. Among the DL networks, the Siamese Convolutional Neural Network (SCNN) is implemented with three branches of ResNet-34 to discriminate between the AD and NC from the Structural Magnetic Resonance Imaging (sMRI). We selected 235 subjects from OASIS dataset. The proposed method achieved the accuracy of 98.72%. The proposed method has the best performance compared with the previous state of the art methods.