Detecção de Covid-19 em imagens radiológicas torácicas através da rede neural convolucional (RNC)

Abstract

The new Covid-19 pandemic, the new coronavirus (SARS-CoV-2), has caused many problems in several areas of society. Among these problems, the high number of tests

carried out, and the lack of tests, especially in small cities, stands out efficiently. Consid- ering this insufficient resource in the identification of Covid-19, the objective of this work

was to propose a solution, in an experimental study, to enable the detection of the new virus using the images of chest radiological medical examinations, which have facilitators in their realization, such as cost and availability, with the help of an Artificial Intelligence, more specifically the Convolutional Neural Network (RNC), which they are usually used

to analyze images. The work consisted of its own architecture and a set of 16,500 im- ages separated into 3 classes (Normal, Covid-19 and Pneumonia), and the results of the

RNC were analyzed and took place on the accuracy of the network algorithms in image recognition disorders, which is higher than 80 %. The technique proved to be promising in meeting this demand, in an experimental context. Therefore, it can be indicated as a complementary instrument in medical diagnosis in the health system for its efficacy and effectiveness, in addition to the benefits and ease for its implementation, however more tests are carried out, with a set of wide variables.