Local Binary Pattern Based COVID-19 Detection Method Using Chest X-Ray Images

COVID-19 is a global pandemic disease that is rapidly spreading around the world. Automatic early diagnosis of COVID-19 with computer-aided tools is crucial for disease treatment and control. In this context, X-ray imaging is an easily accessible and alternative tool in the early diagnosis of COVID-19. However, various lung diseases such as COVID-19, viral pneumonia, bacterial pneumonia are similar to each other and these images may not be distinguished from each other. Thus, the similarity of COVID-19 symptoms to viral pneumonia can lead to misdiagnosis. In this study, the local binary pattern (LBP) based COVID-19 detection method is studied. The textural features are extracted with LBP and supervised learning methods are performed with these features. Different classifiers such as kNN, Naive Bayes, Neural Network, and SVM are used in the training stage and experimental studies are conducted on an openaccess dataset. Performance evaluations of classifiers are made with various performance metrics. As a result of experimental studies conducted in different types and dimensions, over 99% accuracy is achieved with the LBP+SVM method.