AI Models for Accurate Bacterial Pneumonia Diagnosis in Chest X-ray Images

Authors : Cihan Aydin, Hafize Kızılkaya, Muhammet Emin Şahin, Hasan Ulutaş

Pages : 169-177

Doi:10.16919/bozoktip.1593097

View : 48 | Download : 43

Publication Date : 2025-06-15

Article Type : Research Paper

Abstract :Objective: This study aims to contribute to this gap by evaluating the performance of various deep learning models, including a proposed CNN model, ResNet50, and EfficientNetB0, for the detection of bacterial pneumonia from chest X-rays. Material and methods: This study investigates the use of artificial intelligence (AI) in detecting pneumonia from chest X-ray (CXR) images using deep learning techniques, specifically Convolutional Neural Networks (CNN), ResNet50, and EfficientNetB0. Results: A created novel dataset consisting of 1,228 images of bacterial pneumonia and 1,228 images of non-pneumonia cases, is used for model training and evaluation. X-ray images obtained from Yozgat Bozok Medical Faculty are classified by a specialist physician and supplemented with additional images from a publicly available dataset to eliminate class imbalance. Three deep learning models are implemented and evaluated in terms of accuracy, precision, recall, and F1-score. All models achieved an accuracy of 97%, with high performance in detecting both pneumonia and non-pneumonia cases. The Proposed CNN model showed precision and recall values of 1.00 and 0.94 for non-pneumonia and 0.95 and 1.00 for pneumonia detection, respectively. EfficientNetB0 and ResNet50 demonstrated similar robust performance. Conclusion: The results indicate that AI-based models can offer reliable and accurate pneumonia detection, supporting clinical decision-making processes and acting as a valuable second opinion for physicians. These findings highlight the potential of AI in enhancing diagnostic accuracy and efficiency, particularly in resource-limited healthcare settings. Further validation with larger datasets and clinical trials is necessary to confirm the generalizability of these models for widespread clinical use.
Keywords : Derin öğrenme, yapay sinir ağları, pnömoni, sınıflandırma, akciğer grafisi

ORIGINAL ARTICLE URL