Neural Network-based Classification of Alzheimer's Disease using MRI: A Comparative Performance Analysis

Alzheimer's Dementia (AD) is a neurodegenerative disease that is generally found in the elderly. Alzheimer's is one of the main problems in elderly people throughout the world. Early detection of Alzheimer's is critical because it allows for better intervention and management to slow the progression of the disease. Cognitive testing and monitoring behavioral changes are the main screening tools for AD. Brain scans using technologies such as magnetic resonance imaging (MRI) or positron emission tomography (PET) are the gold standard in AD diagnosis. Such imaging can detect structural or metabolic brain changes associated with Alzheimer's. The expertise and experience of medical experts as well as image quality greatly influence detection accuracy. Therefore, in this study, an automatic AD detection method based on MRI images using several neural network architectures is proposed. The proposed system is employed for the classification of non-demented, very mild, mild, and moderate AD. Test results show that 99% accuracy is achieved with the Xception and InceptionResNetV2 architecture. This proposed system is expected to help medical practitioners in the diagnosis of AD.