Using deep learning methods to diagnose COVID-19 disease using computed tomography images

There have been epidemics and pandemics at various times in human history. Covid-19 disease was added to these in 2019 and was accepted as a pandemic in 2020 by the World Health Organization. Considering the risk of transmission and lethal effect of the disease, early diagnosis and treatment of the disease is very important. These effects of the disease have increased the need for computerized support systems. Infected areas caused by the disease in the lungs can be easily seen with computed tomography, can be recognized in computerized support systems and can help healthcare professionals in diagnosis. Today, artificial intelligence serves people in many areas such as health, transportation and education. In this study, an artificial intelligence system that can support experts with deep learning techniques has been designed. Computer-aided systems developed with deep learning are frequently used in the field of health due to the high accuracy and reliability it provides. This study, using lung images obtained with CT, aims to detect Covid-19 disease with deep learning techniques. 751 CT images obtained from 118 Covid-19 patients and 628 CT images obtained from 100 healthy people were included in the study. Images are given as an introduction to both methods proposed in the study. In the first method, images were given as input to the proposed network in addition to AlexNet, VGG-16, VGG-19 and GoogleNet networks, and disease detection was performed. In the second method, images were first given as input to VGG-16, VGG-19, DenseNet-121, ResNet-50 and MobileNet networks and feature extraction was performed from CT images. Among the obtained features, the most significant features were selected by Relief-F feature selection method and disease diagnosis was carried out with XGBoost, AdaBoost, Support Vector Machine and Random Forest classifiers. It is thought that this study will help the early diagnosis of Covid-19 disease, as well as reduce the workload of healthcare professionals while reducing costs with the use of artificial intelligence.