Deep Learning in Image Analysis for COVID-19 Diagnosis: a Survey

Abstract

COVID-19 achieved the highest concentration of confirmed cases in the Americas with a significant impact in Latin America and the Caribbean region, where access to water and sanitation is restricted. In this scenario, we surveyed deep learning techniques applied to extract information from images to detect pneumonia caused by SARS-COV-2, directly assisting health professionals through an automatic case screening. We identify the main public and private image datasets and deep network architectures. Thereby, we identified challenges and research directions. Thus, our goal is to provide a theoretical basis to contribute to the development of computational systems to aid the diagnosis of COVID-19.