Abstract
Alzheimer's disease (AD) is the most prevalent cause of dementia among elderly people. This paper presents a deep learning model based on capsule networks for the classification of AD and Normal Control (NC) using MRI images taken from the ADNI dataset. Our framework introduces several modifications to the original capsule networks in order to significantly reduce the number of trainable parameters so that it can be implemented in computers with limited GPU while obtaining comparable results with the literature. According to the overall results of subject independent and subject dependent experiments, the proposed framework compares favorably with many existing approaches while having fewer parameters and/or using a smaller number of MRI images in AD and NC classes.