Deep Learning-Driven Pneumonia Classification: Leveraging CNN Models for Multi-Cause Diagnosis from Chest X-rays

Abstract:- Pneumonia remains a critical health concern globally, often requiring timely and accurate diagnosis toensure effective treatment. Traditional diagnostic methods, such as chest X-ray interpretation, rely heavily onthe expertise of radiologists and can be subject to variability. With advancements in artificial intelligence,particularly deep learning, it is now possible to automate and enhance diagnostic accuracy using convolutionalneural networks (CNNs). This study aims to develop an efficient deep learning model to identify pneumoniaand classify it into bacterial, viral, and fungal categories using open-source chest X-ray datasets. CNNarchitectures such as ResNet-50 and DenseNet-121 were evaluated for this task. DenseNet-121 achievedsuperior results with an accuracy of 92.1%, demonstrating its suitability for this multi-class classificationproblem. The study emphasizes model interpretability using class activation maps (CAMs) and highlightschallenges related to data imbalance and limited fungal pneumonia samples. The findings suggest that CNN-based models, particularly DenseNet-121, can significantly assist in pneumonia diagnosis and subclassidentification, offering potential for clinical integration in real-time diagnostic systems.Keywords — Pneumonia diagnosis, Convolutional Neural Networks (CNN), DenseNet-121, Chest X-ray classification,Class Activation Maps (CAMs)