Improving Pneumonia Detection in Chest X-Rays: A VGG-16 Framework with Fractional Order Optimization

Imaging the chest using X-rays, commonly referred to as a Chest X-ray (CXR), is essential for identifying and tracking pneumonia and various lung conditions. However, classification frameworks designed to analyze CXR images must be robust enough to handle the complexity and variability present in medical imaging data. In this study, we propose a novel pneumonia detection framework that integrates the Fractional Order Adam (FOAdam) optimizer with the VGG-16 architecture, utilizing advanced optimization techniques based on fractional calculus. By implementing dynamic fractional order scheduling through the Fractional Order Scheduler (FOS), our approach enhances convergence and generalization, effectively addressing the limitations of traditional optimizers such as Stochastic Gradient Descent (SGD) and Adam. Extensive experiments conducted on a publicly available CXR dataset demonstrate the superiority of our method, resulting in significant improvements in accuracy, precision, F1-score, and Matthews Correlation Coefficient (MCC). Specifically, our method achieved an accuracy of 92.60%, a precision of 91.52%, an F1-score of 94.25%, and an MCC of 84.20%. These findings underscore the potential of fractional order optimization in advancing medical image classification. This study not only showcases the effectiveness of FOS-FOAdam in binary classification tasks but also establishes a foundation for applying fractional calculus to a wider range of medical imaging challenges, paving the way for improved diagnostic tools in healthcare.