Automatic Identification of Covid-19 regions on CT-images using Deep Learning
European Journal of Molecular & Clinical Medicine,
2020, Volume 7, Issue 3, Pages 668-676
Abstract
Covid-19 is a contagious respiratory illness caused by a new coronavirus called SARS-COV-2, spreads all around the world and death rate increases at an exponential rate. Covid-19 can be diagnosed either by laboratory base approaches such as nucleic acid testing, antigens test and serology (antibody) tests or by medical imaging tools such as X-ray and Computed Tomography (CT). RT-PCR remains the primary and gold standard for diagnosing Covid-19 but due to shortages of RT-PCR kit, CT images can be used as an alternative early detection toolkit of Covid-19 as a simpler, quicker and more reliable diagnosis of Covid-19. As increase in bandwidth of CT images as well as the new Covid-19 virus consumes a lot of time and workload of the radiologist increases substantially. Deep Learning models can assist the radiologist by learning the features of Covid-19 by their annotated CT images. This paper proposes novel deep learning models for the three main tasks namely 1. Binary classification of Covid-19 2. Automatic lung segmentation and 3. Covid-19 region segmentation. The proposed deep learning models produce an accuracy of 97%, 98% and 99% respectively. The results of the deep learning models show that the models can assist radiologists for quick, accurate and unbiased diagnosis for Covid-19.
Keywords:
- Article View: 113
- PDF Download: 233