Alzheimer's disease Detection using 3D Deep Learning Model

Abstract

Alzheimer’s disease is one of the leading causes of death in the present era of the world. No treatment is available for Alzheimer’s disease after it is in a higher stage. Therefore, it is required to detect disease, when it is in the initial stage. The physical symptoms of the disease cannot be noticed by the patient, in an earlier stage, that’s why researchers proposed deep learning-based solutions to detect it earlier. In this study, I aimed to present an architecture of convolutional neural network (CNN) model for detecting Alzheimer’s disease to handle 2D and 3D data. The proposed 2D model is fast and accurate compared to the AlexNet model. I pre-processed the Alzheimer’s Disease Neuroimaging Initiative (ADNI) image dataset of the patient’s magnetic resonance images (MRIs). After training and testing of the presented model, I obtained 72.13% accuracy to determine AD vs. NC. The proposed 3D model is more accurate compared to one existing research work. ADNI dataset of T1-weighted images had been used and pre-processed for testing of model. After validating the model on that, it gives 96.15% accuracy. Details about the survey of research work and the proposed model architecture are given in this report.