COVID 19 pneumonia analysis from radiology images

COVID19 is a disease that appeared in China,Wuhan in end of the 2019,December.It became a pandemic within a few months. The virus that causes the disease is SARS-CoV-2, which belongs to the SARS virus family. The main symptoms of the disease are high fever, weakness, shortness of breath and cough. In the later stages of the disease, results that increase mortality such as various digestive system disorders, heart diseases and pneumonia can be seen. Radiological images are used during the COVID19 pandemic period to examine whether the disease causes pneumonia and to monitor the situation if pneumonia has occurred in the lungs. Radiological images can be explained under several headings; computed tomography images, magnetic resonance images, X-ray images and positron emission tomography images. Computed tomography can generally be used to examine various regions such as bone tissue, vascular tissues, heart tissue and lung tissues. During the COVID19 pandemic, it is one of the most used methods for detecting pneumonia. The main reasons for this can be counted as offering high-resolution images, getting results in a short time, and providing detailed information because it offers multi-layered images. Computerized tomography images can be used to detect lesions as well as to follow the course of the disease on the lungs. In this study, 12967 images were trained and tested by VGG16, ResNet50, ResNet50v2 and ResNet50v2 architectures with added layers. Class activation was visualized on images with the ResNet50v2 model with the trained layer added. According to the results of the tests, the ResNet50v2 model with the added layer gave 97.83% sensitivity, 99.47% specificity and 99.53% test accuracy.