This work investigates improvements in pneumonia detection by utilizing deep learning models with chest radiograph images, focusing on both accuracy and interpretability. The study highlights the feasibility of computer-aided diagnosis (CAD) systems that rely on deep learning-based feature extraction and classification models to provide accurate, interpretable, and efficient pneumonia diagnosis. Such systems have significant implications for both high-resource health care settings, where precision and speed are critical, and low-resource environments, where reliable automated tools can support limited medical expertise. In this implementation, a differential evolution (DE) algorithm is employed to enhance feature selection for pneumonia detection from chest X-ray images. The primary objective is to optimize a high-dimensional dataset by identifying the most informative features, thereby improving the overall performance of classification models. The proposed two-stage feature selection method, which incorporates the Voting Differential Evolution (VDE) algorithm, achieved an impressive 98.67% accuracy while successfully reducing feature dimensions by 19.93%.