Classification of Chest X-ray COVID-19 Images Using the Local Binary Pattern Feature Extraction Method

Abstract: Background and Purpose: COVID-19, which started in December 2019, caused great loss of life and economic losses. In order to early diagnosis of the disease is so important, we worked on diagnostic systems supported by machine learning methods to assist experts in this field and automatically identify COVID-19 in chest radiology images. Materials and Methods: In this study dataset consists of a total of 15153 X-ray images for 4961 patient cases in three classes: Viral Pneumonia, Normal and COVID-19. Firstly, the dataset is preprocessed and given as input to the Cubic Support Vector Machine (Cubic SVM), Linear Discriminant (LD), Quadratic Discriminant (QD), Ensemble, Kernel Naive Bayes (KNB), K-Nearest Neighbor Weighted (KNN Weighted) classification methods. In order to improve the obtained performance criteria results, feature extraction is applied with the help of the Local Binary Pattern (LBP) texture operator. Results: The analysis results obtained after applying LBP feature extraction to the input data showed an increase compared to the analysis results obtained without feature extraction. In these two different analyzes, the highest estimation accuracy is seen in the Cubic SVM method. It observed that the estimation accuracy increased from 94.1% to 98.05% and other performance criteria improved.