COVID-19 detection with image processing and deep learning from lungs computed tomography

The novel coronavirus disease (COVID-19) is an epidemic disease caused by the SARS-CoV-2 virus. Due to the high contagiousness of the disease, infected individuals must be identified and isolated quickly to control the disease. The most preferred method for detecting the disease is reverse transcription-polymerase chain reaction (RT-PCR) tests. The length of the results of these tests and the fact that the percentage of success can differ according to the stages of the disease stand out as important disadvantages. Medical radiological imaging methods are also used to diagnose COVID-19 quickly and accurately in the early stages of the disease. These methods have less risk of infection as they require less contact with the patient. Studies in the literature that target the detection of COVID-19 with deep learning-based approaches, especially through X-Ray (X-Ray) and Computed Tomography (CT) images attract intensive attention. In this study, a dataset including lung CT images was created. On this dataset, the detection of COVID-19 was carried out using deep learning-based architectures. Two different strategies were followed when using deep learning architectures. In the first strategy, the effect of network depth on performance was examined using Basic-CNN and VGG16 architectures. In the second strategy, the effect of the learning transfer method on the performance was examined using VGG19, MobileNet and DenseNet. As a result of the studies, it has been observed that the DenseNet201 architecture has the highest performance with a test accuracy of 0,99.