Pneumonia Detection Using X-ray Images Based On Deep Learning Technology

DOI: https://doie.org/10.10399/JBSE.2025462122

Omkar Kasulwar , Bhakti Sonawane

Keywords:

Pneumonia Detection, Deep Learning, Convolutional Neural Network, Pre-trained Models, Medical Imaging, Chest X-ray.

Abstract:

Pneumonia is a dominant worldwide cause of morbidity and mortality, mainly among vulnerable patient groups like the elderly and children. In many diagnostic protocols, chest X-ray is in regular use, although precise interpretation must often be based on the evaluation of experienced radiologists, giving rise to gaps in the efficiency of diagnosis in a timely fashion, particularly under resource-limited conditions. Modern developments in the field of deep learning, mainly convolutional neural networks (CNNs), hold great potential to automate pneumonia diagnoses from radiologic images. This work compares the performance of some pre-trained CNN models—LeNet, GoogleNet, VGG16, and Inception V3—against a manually developed Manual Net model. The models were tested on an open-source chest X-ray data set, where performance was evaluated in terms of accuracy, precision, recall, and F1-score. The findings show that although pre-trained networks provide robust baseline performance, the Manual Net architecture provides a potential lightweight alternative with lower computational demands, making it deployable in low-resource settings. The work also discusses interpretability, training efficiency, and dataset issues. Through this work, we shed light on the choice and optimization of CNN models for real-world pneumonia detection, moving towards more accessible and accurate diagnostic tools.

PDF