Whale Optimized Deep Generative Adversarial Network Based Alzheimer's Stages Detection Using 3D MRI Brain Neuroimaging

Abstract

Alzheimer's Disease (AD), a common, chronic neurodegenerative condition, is characterized by the loss of neurons and synapses in the cerebral cortex and specific subcortical regions. According to claims from a recent study, AD has a 20% misdiagnosis rate. Therefore, it is essential to create a useful tool to recognize the stages of AD with a lower prediction error rate to reduce misdiagnosis. Hence proposed a model called Whale-Optimized Deep Generative Adversarial Network (WODGAN). A generator plus a discriminator make up the model. The discriminator trains the model using real images; The generator creates synthetic images using noise and random selection. The discriminator goes through some processes to improve image quality, including Adaptive Histogram Equalization (AHE) and Adaptive Filtering (AF) approaches. Fuzzy feature extraction techniques are used to accurately segregate biomarker regions from brain MRI scans depending on AD pathology. The model uses Hilbert-Schmidt Independence Criterion Lasso (HSICL) to discover optimized biomarker features to combat overfitting. Before training, the discriminator can tell actual photos from artificial ones. The Whale Optimizer (WO) is used during training to improve network efficiency and lower prediction errors. The numerical results show a high accuracy of 99.93% in AD stage recognition.