Optimized VGG16 for Multi-class Classification of Alzheimer's Disease

Alzheimer’s disease (AD) is a severe neurological disorder that leads to memory loss and other cognitive impairments, ultimately resulting in death. With the advancement of deep learning, several researchers implemented a deep learning (DL)-based classification system for diagnosing such disease. However, determining the proper setting of the parameters, like learning rate and the number of dense units in the classification model is essential to achieve the best performance. Although manual tuning and grid search methods may be used, they are often time-consuming and cannot achieve the optimal model’s design. Accordingly, this paper proposed an improved model of the Visual Geometry Group (VGG) using the Particle Swarm Optimization (PSO) method to enhance classification-based diagnosis of Alzheimer's disease. The proposed system adjusted the important parameters in the VGG model to enhance early detection and multi-class classification of Alzheimer's disease. The effectiveness of the proposed model was evaluated in different scenarios, where the model's performance was assessed before and after applying the optimization process. Additionally, its performance was compared using different optimization algorithms such as Genetic Algorithm (GA), and Political Optimization (PO). Finally, MRI images were used to evaluate the proposed system against known models. The results showed that the proposed VGG16 optimization model outperformed when tuned using PSO with an accuracy of 99.14%.