Comparative Analysis of Deep Learning Architectures for the Identification of Pneumonia in Chest Radiographs

Pneumonia has been noted as one of the major contributing diseases with regard to disease burden and mortality rate and timely diagnose is crucial. Conventional Chest X-ray (CXR) imaging remains a basic approach to identifying pneumonia and targeting critical lung abnormalities. However, the ability to interpret CXR can be limited by issues related to the availability of a specialized radiologist, especially where the setting is impoverished. This research employs deep learning strategies to increase the specificity of the diagnosis of pneumonia through CXR images. Different CXR models that are based on CNN like ConvNext, Vision Transformers (VTs), Xception, VGG16 were trained and tested with the datasets. The assessments of these models were done according to accuracy, sensitivity and specificity. The study's goal is to show how applying deep learning models to CXR analysis could improve pneumonia diagnosis as a faster and more accurate diagnostic technique. Following evaluation of these models, it is significant to highlight that VGG 16 models had higher accuracy rates of 93.11%, which will be useful when using CXR pictures to diagnose pneumonia.