Multi-class Pneumonia Disease Detection Using Machine Learning and Deep Learning on Chest X-ray Images

Abstract

Pneumonia remains a leading global health concern, necessitating rapid and precise diagnostic tools to reduce mor- tality. While manual interpretation of chest X-rays is expert- dependent and time-intensive, automated systems offer a scalable alternative. This paper proposes a hybrid diagnostic framework for multi-class pneumonia detection using a combination of Machine Learning (ML) optimization and Deep Learning (DL) architectures. The system employs a Convolutional Neural Net- work (CNN) to classify radiographs into three distinct categories: Normal, Viral Pneumonia, and Bacterial Pneumonia. To ensure clinical transparency, the model integrates Gradient-weighted Class Activation Mapping (Grad-CAM) to provide visual in- terpretability by localizing pathological markers. Experimental results demonstrate a high classification accuracy of 94%, with the interpretability layer successfully aligning AI predictions with radiological features. This integrated ML-DL approach provides a robust, explainable, and consistent diagnostic aid, particularly suited for resource-constrained medical environments.