AlzNET: Computational Clinical Decision Support for Diagnosing Alzheimer's Disease from MRI Images

Globally Alzheimer's disease (AD) has been recognized as the primary cause of dementia which may eventually lead to fatality. It is a neurodegenerative disorder that causes adverse effects like the death of brain cells and brain atrophy. The deep learning approach has been employed in every field due to its efficacy in producing desired results, especially in medical imaging domain. In this paper, we developed a customized deep learning-based computational model called AlzNET for the classification of Alzheimer's disease. The data set used in this study contains Magnetic Resonance Imaging (MRI) brain scan images of normal and AD-affected people of different stages. The AlzNET model categorizes the MRI images into one of four categories including mild-demented, moderate-demented., non-demented., and very mild-demented based on the severity. To construct the overall framework for our proposed model, we use the Convolutional Neural Network (CNN) that reports an accuracy of 0.821. The images were transformed before feeding into the framework to process the MRI faster. Experimental results demonstrates that the proposed AlzNET computational model outperforms other popular deep neural networks family including VGG19, ResNet 101, DenseNet169, and InceptionV3.