COVID-XR: A Web Management Platform for Coronavirus Detection on X-ray Chest Images

OROZCO, CARLOS ISMAEL; XAMENA, EDUARDO; CRISTIAN A. MARTÍNEZ; DIEGO A. RODRIGUEZ

IEEE LATIN AMERICA TRANSACTIONS

IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC

Lugar: New York; Año: 2021 vol. 19 p. 1033 - 1040

1548-0992

COVID-19 is an infectious disease caused by the SARS-CoV-2 virus. Its symptoms are similar to those of the common flu, including fever, cough, dyspnea, myalgia, and fatigue. Due to its rapid expansion globally, the World Health Organization (OMS) declared it a pandemic. The molecular test commonly used worldwide for direct detection of the virus is the RT-PCR test but it takes time to process and the materials used are scarce. In this work we propose: (a) The design and implementation of a deep neural network architecture for the detection of patients with COVID-19 using as input X-ray images of the chest; the architecture is made up of a feature extraction phase, that is, a pre-trained model VGG16 extracts the features of the image; then in the second phase, a multilayer neural network classifies into one of two particular classes (1: COVID, 0: NO COVID). (b) The implementation of a Web platform that allows interested people to use our architecture in a clear, simple and transparent way. The deep learning algorithm was implemented in Python with specific libraries for the design of neural networks, while the Web platform was implemented in PHP using the Laravel framework and MySQL database. We evaluate the performance of our proposal using the sensitivity, specificity and area under the curve (AUC) evaluation metrics, obtaining good results in very short computational times.