TRANSFORMER BASED COVID-19 DETECTION USING CHEST X-RAYS

Abstract

Covid-19 has affected millions globally, leading to substantial illness and mortality. Chest X-rays serve as a rapid and effective means of tracking the progression of Covid-19. However, diagnosing Covid-19 from a chest X-ray can be complex, and even skilled radiologists may not always provide a conclusive diagnosis. In our research, we utilized a dataset comprising X-ray images of Covid-19, lung opacity, viral pneumonia, and healthy patients to assess the efficacy of various vision transformer-based models. A modified version of the Swin Transformer achieved an accuracy of 98.9% and a precision of 99.2% on Covid-19 images in a four-way classification task. Our findings are competitive with cutting-edge techniques for diagnosing Covid-19. This method could aid healthcare professionals in screening patients for Covid-19, thereby enabling quicker treatment and improved health outcomes for those affected by the virus.