Determination of Covid-19 Possible Cases by Using Deep Learning Techniques

Authors : Çinare OĞUZ, Mete YAĞANOĞLU

Pages : 1-11

Doi:10.16984/saufenbilder.774435

View : 14 | Download : 9

Publication Date : 2021-02-01

Article Type : Research Paper

Abstract :A large number of cases have been identified in the world with the emergence of COVID-19 and the rapid spread of the virus. Thousands of people have died due to COVID-19. This very spreading virus may result in serious consequnces including pneumonia, kidney failure acute respiratory infection. It can even cause death in severe cases. Therefore, early diagnosis is vital. Due to the limited number of COVID-19 test kits, one of the first diagnostic techniques in suspected COVID-19 patients is to have Thorax Computed Tomography insert ignore into journalissuearticles values(CT); applied to individuals with suspected COVID-19 cases when it is not possible to administer these test kits. In this study, it was aimed to analyze the CT images automatically and to direct probable COVID-19 cases to PCR test quickly in order to make quick controls and ease the burden of healthcare workers. ResNet-50 and Alexnet deep learning techniques were used in the extraction of deep features. Their performance was measured using Support Vector Machines insert ignore into journalissuearticles values(SVM);, Nearest neighbor algorithm insert ignore into journalissuearticles values(KNN);, Linear Discrimination Analysis insert ignore into journalissuearticles values(LDA);, Decision trees, Random forest insert ignore into journalissuearticles values(RF); and Naive Bayes methods as the methods of classification. The best results were obtained with ResNet-50 and SVM classification methods. The success rate was found as 95.18%.
Keywords : Resnet 50, Alexnet, Deep Learning, COVID 19, Classification