Chest X-Ray Disease Classification using ML

Abstract

This paper reviews recent advancements in deep learning-based medical image classification, emphasizing chest X-rays for COVID-19 and pneumonia detection. The objective is to analyze research trends and evaluate how evolving architectures enhance diagnostic performance. Ten key studies published between 2022 and 2025 were examined to identify a transition from conventional transfer learning models to hybrid and attention-driven networks. Early research focused on optimizing pre-trained architectures such as VGG-16 and ResNet50 through feature refinement and fine-tuning, whereas recent approaches incorporate conditional cascaded networks, attention capsule frameworks, and self-supervised learning to address challenges associated with limited medical datasets. The comparative analysis highlights that multi-stage and hybrid deep learning models consistently achieve superior accuracy and generalization. The review concludes that integrating multiple optimization and feature extraction strategies is critical for improving reliability and interpretability in automated radiographic diagnosis. These insights contribute to ongoing efforts toward developing efficient, data-adaptive AI systems for clinical applications.