Automatic X-ray COVID-19 Lung Image Classification System based on Multi-Level Thresholding and Support Vector Machine

Aboul Ella Hassanien; Lamia Nabil Mahdy; Kadry Ali Ezzat; Haytham H. Elmousalami; Hassan Aboul Ella

In the end of March 2020, more than +724000 confirmed cases of COVID 19 and more than +34000 deaths are exist globally where the humanity in our plant currently lives in COVID 19 pandemic. Isolation and social distance are temporary unpractical solution against fighting COVID-19. Unfortunately, a coronavirus vaccine is expected to take at least 18 months if it works at all. Moreover, COVID -19 pandemics can mutate into a more aggressive form [8] . Therefore, this paper presents a novel COVID-19 detecting methodology based on multilevel thresholding and SVM for X-ray images. The technique is useful for the clinical practitioner for early detection of COVID-19 infected patient. The model presents high accuracy where the average sensitivity, specificity, and accuracy of the lung classification were 95.76%, 99.7%, and 97.48%, respectively. Machine learning algorithms can present high performance in terms of accuracy and computational complexity [26] . Therefore, the future research may be based on using a modified model of optimized SVM or hybrid ML models. Moreover, larger data set can be providing higher model generalization.