Classification of Lung CT-Scan Images for Covid-19 Detection Using Texture Feature Extraction and Naive Bayes Algorithm

Abstract

Covid-19 diagnostic testing is divided into 2 approaches, namely laboratory-based approaches such as Swab Tests, PCR Tests, and Rapid Tests. To minimize transmission, a second approach is used, namely a non-laboratory approach that uses imaging diagnostic tools such as x-rays and Computed Tomography (CT). It is hoped that the use of this non-laboratory approach can be an alternative for detecting the Covid-19 virus and will also minimize the spread of the covid virus because at the time the diagnosis process takes place there is little direct contact with patients. Chest CT scan has a high sensitivity for diagnosing Covid-19. A CT scan uses the transmission of x-rays through the patient's chest, which is then processed with high-resolution medical images. In this study, the classification of Covid-19 images through CT-Scan images of the lungs using texture feature extraction and the naive Bayes classifier algorithm is a system that enters information in the form of input images and compares them with images in the database. The data set used is 200 CT-Scan images of Covid-19 lungs and 200 CT-Scan images of Non-Covid-19 lungs. The training data uses 150 Covid-19 images and 150 Non-Covid-19 images, while the rest is used as test data. The first method used is feature extraction, namely mean intensity, standard deviation, skewness, energy, entropy, and smoothness. After successfully obtaining the feature, it is calculated using the Naive Bayes classifier algorithm. The results obtained from this study get an Accuracy rate of 54%, Precision of 55.88%.