Detecting covid-19 from computed tomography images using deep learning methods

A new coronavirus disease Covid-19 has been recorded in Wuhan, China since late December 2019 and later became a worldwide pandemic. It can result in death as a result of covid-19's damage to the alveoli and progressive respiratory failure. Although transcription polymerase chain reaction (RT-PCR) is the gold standard used for clinical diagnosis, tests can produce false negatives. In addition, in the event of a pandemic, the lack of RT-PCR testing resources may delay diagnosis and treatment. Under these circumstances, Computed Tomography (CT) scans have become a valuable tool for both early diagnosis and prognosis of Covid-19 patients. Recently, many studies developed with deep learning techniques have been proposed to facilitate the diagnosis of Covid-19 in CT scans and to assist healthcare professionals. This thesis focuses on distinguishing COVİD-19 from non-Covid-19 cases using DenseNet121, one of the deep learning techniques, with the CT dataset colored using the DeOldify library. At the end of our study, an accuracy of 0.98 was obtained.