Early Detection of Alzheimer's Disease Using Deep Learning

摘要

Using a combination of methods from image processing, signal processing and deep learning, we aim to develop a model to predict whether or not a patient will develop symptomatic Alzheimer's disease using Diffusion MRI (dMRI) imaging data. We first propose a 3D multichannel convolutional neural network (CNN) architecture to distinguish patients with Alzheimer's from normal controls, then propose an extension of our architecture to incorporate multiple scans from a patient's history to improve classification accuracy and predict future prognosis. Finally, we discuss methods for performing data augmentation to add diversity and robustness to our unique and comparatively small dataset.