Automatic Detection of Covid-19 from Colorized CT Images using Deep Learning

Abstract

Since late December 2019, a new coronavirus outbreak of Covid-19 has been recorded in Wuhan, China, and then became epidemic all over the world. The onset of Covid-19 can result in death as a result of important alveolar damage and advancing respiratory insufficiency. Though the transcriptionpolymerase chain reaction (RT-PCR) used for clinical determination is the gold standard, tests can obvious false negatives. In addition, in the pandemic situation, insufficient RT-PCR test resources may delay diagnosis and treatment. Under these circumstances, Computed Tomography (CT) scans have become a precious vehicle for both early diagnosis and prognosis of Covid-19 patients. Recently, many studies developed with deep learning techniques have been proffered to facilitate the diagnosis of Covid-19 in CT scans and to assist healthcare professionals. The purpose of this article is to first create a mixed dataset by coloring some of the CT images with the DeOldify method in order to make a more right performance and then to detect COVID-19 cases using DenseNet121, one of the deep learning (DL) techniques. © 2021 IEEE.