Enhancing Alzheimer’s Disease Classification by Employing Deep Learning and Optimization

Abstract

Alzheimer's Disease (AD) is a progressive neurological disorder that leads to the deterioration of memory and cognitive abilities due to damage brain's nerve cells. Deep Learning (DL) techniques can provide effective AD classification by using medical imaging data. In this study, a DL technique based on Convolutional Neural Networks (CNN) is established. Keras-Tuner Optimization (KTO) is applied since it is difficult to propose a CNN with the suitable architecture. The aim of the proposed Optimized CNN (OCNN) model is to classify AD into four groups Mild Dementia (MD), Moderate Dementia (MoD), Non-Dementia (ND), and Very Mild Dementia (VMD). Two datasets are utilized here namely: Best Alzheimer Magnetic Resonance Imaging (BA-MRI) and Alzheimer’s Disease Magnetic Resonance Imaging (AD-MRI). The Kaggle platform is the source and collection point for both datasets. After extensive implementations and OCNN models training, high classification accuracies of 92.44% and 93.17% are achieved for the AD-MRI and BA-MRI datasets, respectively.