Kadiroğlu, ZehraDeniz, ErkanŞenyiğit, Abdurrahman2024-03-252024-03-252024Kadiroğlu, Z., Deniz, E. ve Şenyiğit, A. (2024). A comparison of deep learning models for pneumonia detection from chest x-ray images. Journal of the Faculty of Engineering and Architecture of Gazi University, 39(2), 729-740.1300-1884https://dergipark.org.tr/tr/pub/gazimmfd/issue/80437/1204092https://hdl.handle.net/11468/13689Pneumonia is one of the acute lower respiratory tract diseases that can cause severe inflammation of the lung tissue. Although chest X-ray (CXR) is the most common clinical method for diagnosing pneumonia due to its low cost and ease of access, diagnosing pneumonia from CXR images is a difficult task even for specialist radiologists. It has been shown in the literature that deep learning-based image processing is effective in the automatic diagnosis of pneumonia. In conclusion, deep learning-based approaches were used in this study to classify pneumonia and healthy CXR images. These approaches are deep feature extraction, fine-tuning of pre-trained Convolutional Neural Networks (CNN), and end-to-end training of an enhanced ESA model. For deep feature extraction and transfer learning, 10 different pre-trained deep CNN models (AlexNet, ResNet50, DenseNet201, VGG16, VGG19, DarkNet53, ShuffleNet, Squeezenet, NASNetMobile and MobileNetV2) were used. Support Vector Machines (SVM), k Nearest Neighbor (kNN), Random Forest (RF) classifiers are used to classify deep features. The success of the fine-tuned AlexNet model produced an accuracy score of 98.50%, the highest of all results achieved. The end-to-end training of the developed ESA model yielded 96.75% results. The data set used in this study consists of Pneumonia and healthy CXR images obtained from Dicle University Medical Faculty Pulmonary Diseases and Tuberculosis clinic, intensive care unit and pulmonary outpatients’ clinic.eninfo:eu-repo/semantics/openAccessChest x-ray imagesConvolutional neural networksDeep feature extractionPneumonia detectionTransfer learningArticle392729740WOS:0011179615000072-s2.0-85180156403124650610.17341/gazimmfd.1204092Q2N/A