Multi-class Alzheimer's Disease Stage Diagnosis using Deep Learning Techniques

Abstract

Alzheimer's disease (AD) is a type of dementia that affects thinking, behavior, and memory. Eventually, symptoms become serious enough to interfere with daily activities of a patient. Older people are typically affected by AD. The complexity of the brain's structure and functioning makes early AD diagnosis difficult, despite the fact that research on the disease has grown significantly. Inadequate datasets on AD are a major barrier to furthering research into the disease's diagnosis. On the other hand, stages of AD classification remain a challenge to this date for the Deep Learning (DL)-based techniques previously employed in many studies. This study solves the challenge of insufficient data on the AD dataset and the poor accuracy of multi-class stage diagnosis by proposing a DL framework based on a transfer learning approach. A total of five different transfer learning models were trained, and from the evaluation, it has been clear that almost all the models performed greatly compared to the best performing model that was proposed, which was Inception V3with an overall accuracy of 98.58%.