A comprehensive Analysis on Diagnosis of Alzheimer Disease Using Generative Adversarial Network

Chronic stress in daily life causes physiological changes, including melancholy, anxiety, and cognitive deterioration. AD, the most common type of dementia in older people, is characterized by gradual memory loss, disorientation, and reduced language and decision-making skills. The major research databases (including Scopus, Web of Science, and IEEE Explore) are used to search for articles on generative adversarial networks (GANs) to analyze AD. To identify similar articles, search engines employ keywords, like “Alzheimer’s,” “detection,” “Magnetic Resonance Imaging (MRI),” “GAN,” and so on. Following a comprehensive screening method, 532 abstracts are gathered from multiple databases, and 19 full-text articles are selected for review. The analysis focuses on articles that use preexisting AD image datasets, implementing Machine Learning (ML) and GAN. Further, Deep Convolutional GANs (DCGANs), Wasserstein GANs (WGANs), Conditional GANs (CGAN), and Auxiliary Classifier GANs (ACGANs) are being investigated. Performance metrics such as sensitivity, specificity, F1-score, Recall, Precision, AUC, loss functions, and accuracy are discussed. This study consistently shows that GANs excel over other ML algorithms in terms of diagnostic accuracy. However, difficulties still exist in integrating multi-modal imaging data and attaining generalization across different datasets. Finally, this thorough work sheds light on the use of GANs for the systematic and non-invasive detection of AD using neuroimage analysis.