Seleção de atributos e classificação de imagens radiográficas em paciente com COVID-19

In December 2019 a disease appeared caused by a new type of coronavirus, Covid-19. The disease has become a pandemic, to date. One of the ways to avoid contamination is through social isolation and especially isolation and rapid diagnosis of the patient. For the diagnosis, it is necessary to perform the RT-PCR exam, through a blood sample, but as one of the characteristics of the disease is the damage to the lungs, it is possible to detect it through CT scans and X-rays. the similarity of the images of the results of patients diagnosed with Covid-19 and the results of patients who have other diseases. Thus, we used the K-means technique to differentiate radiographic images of patients with Covid-19 and those without the disease. Analyzing the Haralick texture descriptors without individual parameters, we observed that the highest hit rate occurred for the Entropy Difference descriptor with 95.83% hits, followed by the Inverse Moment Difference descriptor with 94.44% hits.