University of Twente Student Theses


Feature evaluation for detection of COVID-19 in pulmonary CT scan images

Popa, A. (2020) Feature evaluation for detection of COVID-19 in pulmonary CT scan images.

[img] PDF
Abstract:At the beginning of 2020 a novel coronavirus, SARS-CoV-2, has started its spread, which has become a global issue. The current method of detecting the infection with the disease (COVID-19) caused by the virus is the transcription polymerase chain reaction (RT-PCR), which is limited by the low detection rate when the viral load is low and the high demand for it, which means that a new detection method is needed. Another method, that has the possibility to detect the signs of COVID-19, is through medical imaging, like computed tomography (CT) scan or X-rays scans of the chest. The method that is presented by this paper uses the patches of the pulmonary 2D CT scans to extract feature vectors and a support vector machine (SVM) classifier to detect the affections caused by the disease. The aim is to see which characteristics of the affection, namely color, texture, position or the combination of all three of them, are better suited in detecting the signs of the disease. In order to evaluate the results different values for regularization and class weights parameters are used and the precision, recall and Matthews correlation coefficient are computed. The dataset contains 100 CT scan images of patients that show signs of the disease. The results show that the best single feature vector that can be used to detect the affections caused by the disease is the color, while the best results were achieved by the combination of all three features.
Item Type:Essay (Bachelor)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science BSc (56964)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page