Identification of COVID-19 Based on Features Texture Histogram and Gray Level Co-Occurrence Matrix (GLCM) Using K-Means Clustering Methods in Chest X-Ray Digital Images

Abstract

Since the last five years of the COVID-19 outbreak, radiological images, such as CT-Scan and Chest X-Ray (CXR), have become essential in diagnosing this disease. However, limited access to facilities such as CT-Scanners and RT-PCR makes CXR images the primary method for COVID-19 testing. This research aims to improve the accuracy of CXR images in identifying COVID-19 patients based on the texture features: histogram and Gray Level Co-occurrence Matrix (GLCM), using the K-Means Clustering method. This study utilized 150 CXR images, including 75 COVID-19 patients confirmed by RT-PCR tests, and 75 patients with negative cases. The method used were consisted of pre-processing, and texture feature extraction with the seven most influential attributes based on gained information (histogram: standard deviation, entropy, skewness, kurtosis, and GLCM: correlation, energy, homogeneity), as well as classification using K-Means clustering methods. The results showed that the classification’s accuracy, sensitivity, and specification are 92%, 91%, and 93%, respectively. This image processing technique is a promising as well as a complementary tool in diagnosing COVID-19 cases, based on CXR images with lower costs and more reliable results.