Alzheimer’s disease is a major public health problem worldwide. Timely and accurate diagnosis is crucial for effective treatment and patient care. Recent progress in deep learning holds promise for enhancing the detection of Alzheimer's disease through the analysis of neuroimaging data, including magnetic resonance imaging (MRI) and positron emission tomography (PET). The deep learning models examined in this study include convolutional neural networks (CNNs). We rigorously analyze the effects of different optimizer algorithms, including Adam, RMSprop, and stochastic gradient descent (SGD). Additionally, we explore the influence of batch sizes during the model training process. One of the key contributions of this research is the in-depth examination of batch size selection in the context of AD detection. We demonstrate how batch size affects training dynamics, convergence speed, and memory requirements, shedding light on the trade-offs associated with different batch sizes. We assess the model’s performance using robust evaluation metrics, including accuracy, sensitivity, specificity, and the area under the receiver operating characteristic curve (AUC-ROC). In this paper, results showcase that no one-size-fits-all approach exists, as the optimal combination of model, optimizer, and batch size depends on the specific problem domain and dataset characteristics. This research paper serves as a reference point for researchers and practitioners seeking to maximize the accuracy of their deep-learning models. By systematically exploring the interplay between different factors in model optimization, we offer a nuanced perspective on achieving superior model performance across diverse tasks and datasets.