Potential use of Convolutional Neural Networks in Alzheimer’s Detection

Abstract

Worldwide, Alzheimer’s disease (AD) affects one in nine individuals over the age of 65. AD is a progressive disease that kills neurons in parts of the brain related to mental function, causing memory loss and the irreversible decline of communication or problem solving abilities. Early diagnosis is important for those with AD, as it allows them to make informed decisions regarding treatment options and finances prior to significant mental decay. Although there doesn’t exist a cure, different forms of treatment have been shown to slow disease progression, most effective at early stages (Bush 2). The cause of the disease is not yet known, but early symptoms, such as protein buildup and shrinkage in the brain, occur as much as fifteen years before signs of cognitive decline (Scheltens 5). By combining machine learning technology with brain imaging, it should be possible to quickly and effectively detect AD early in its development. The goal of this work is to train multiple convolutional neural networks (CNNs) off of MRI brain scans to detect the severity of AD, and then see if they can outcompete alternative technologies.