Transfer Learning-Based Approach for Early Detection of Alzheimer's Disease

Abstract

Alzheimer's disease is one of the world's main health concerns today. People with Alzheimer's disease who are diagnosed early have the best chance of receiving effective therapy. It's critical to catch the sickness as early as possible. Magnetic resonance imaging is one way to define Alzheimer's disease by finding structural abnormalities in the brain (MRI). We propose that machine learning, specifically trained convolutional neural networks (CNNs) with transfer learning capable of making predictions about similar brain imagery, can aid in early detection. CNN enables the extraction of MRI properties and classification as Alzheimer's disease or normal brain. We used the VGG19 architecture to categorize patients as having no signs of Alzheimer's disease or having signs of very mild, mild, or moderate Alzheimer's disease. Based on a transfer learning methodology, this method correctly classifies MRI images into four phases of Alzheimer's disease with an accuracy of 85 percent.