Alzheimer’s Disease Detection and Severity Classification using Deep Learning Models

Abstract

Alzheimer's disease is a neurodegenerative disease, and its diagnosis, especially in the initial stages, is extremely tough. The recent breakthroughs in deep learning techniques have generated a mania in research activities for better diagnosis with the use of neuroimaging and medical data. The three important facets of this comprehensive survey are, a) types of data used in research for AD, including MRI, PET and multimodal dataset, b) a range of deep learning techniques including CNNs, RNNs, Autoencoders, and Transformers, c) the effectiveness of these techniques in terms of accuracy, sensitivity, and specificity for the aforementioned datasets. The survey involves a comparison and analysis of various results obtained from benchmark datasets such as ADNI and OASIS, and indicates the state of research in this area, future research, and trends in this regard in a manner that would further enhance the use of Deep Learning in AD diagnosis.