Comparison of the Performance of Convolutional Neural Network (CNN) Architectures in Detecting Pneumonia with Chest X-ray Image Data

Acute lung infections called pneumonia are caused by bacteria, viruses, or fungi, which cause inflammation of the air sacs and pleural effusion. To prevent pneumonia, early detection is needed to reduce the death rate due to pneumonia. This research compares the performance of Convolutional Neural Network (CNN) models such as ResNet50, MobileNetV2, DenseNet169, EfficientNet-B0, and Xception when using chest x-ray data to detect pneumonia. The research begins with data collection, image preprocessing, model training, testing, validation, and model evaluation. This study used 5,856 chest X-ray images that were classified as pneumonia and normal. The results show that MobileNetV2 is the best of all models with a training accuracy of 0.91%, testing accuracy of 0.88%, and validation accuracy of 0.93%. DenseNet ranked second with training accuracy of 0.91%, measurement accuracy of 0.88%, and validation accuracy of 0.81%. Third, Xception is in third place with training accuracy of 0.90%, testing accuracy of 0.86%, and validation accuracy of 0.81%.