Abstract
Alzheimer's disease (AD) is a neurodegenerative disease that causes damage associated with memory and thinking, causing a gradual decline in judgment, reasoning and learning. One of the ways to aid in the diagnosis of AD is the analysis of structural magnetic resonance (MR) images of the patient. Recent studies use computer vision, image processing and machine learning techniques to help diagnose AD. This work aims to extract features from MR images with stacked autoencoders and stacked convolutional autoencoders, and classify MR images via XGBoost into cognitively normal (CN) and AD classes. This work also compares Mean Squared Error (MSE) and Structural Similarity Index (SSIM) metrics as loss functions in autoencoders.