ABSTRACT
Alzheimer’s is a brain disease in which the ability to think, remember, and behave significantly decreases. Using different approaches and algorithms to distinguish between different stages of Alzheimer’s disease, neuroimaging data has been used to extract patterns associated with varying stages of Alzheimer’s disease. However, because the brain patterns of older adults and people in different stages of Alzheimer’s disease are similar, researchers have had difficulty classifying them. This chapter proposes an improved residual neural network based on the ResNet-50 network, called the ResNetF network. Moreover, rectified linear unit (ReLU) has replaced the activation function (leaky ReLU) because ReLU takes the negative parts of the input and reduces them to zero while preserving the positive parts. These negative inputs may contain useful feature information that can be used to develop high-level discriminative features. Experimental results show that the modified residual network performs excellently, indicating the robustness and superiority of our model. The Alzheimer’s brain magnetic resonance imaging dataset was obtained from the open-access section of the Kaggle website. The proposed method successfully classifies the four stages of Alzheimer’s disease with an accuracy of 97.49% and 98% for precision, recall, and F1 score.
