COVID-19 Signs Detection in Chest Radiographs Using Convolutional Neural Networks

COVID-19 pandemic that affected the entire world since late 2019 and the need to collaborate with the healthcare system gave rise to this article. Early diagnosis of the disease caused by the coronavirus is crucial for the treatment and control of this type of illness. In this context, chest radiography plays an important role as an alternative test to confirm or rule out an infected person; precisely, this work aims to analyze and develop a software prototype for COVID-19 signs recognition in chest radiographs, based on image processing using convolutional neural network model. Proposal is based on CRISP-ML methodology, following its phases of understanding the business and data, adapting the latter, generating the model and then evaluating it; experimentation and analysis of the behavior of the network were carried out by training it using different publicly available datasets. Experimental results demonstrate the proposed prototype effectiveness and limitations, with classification accuracy close to 80%.