Análise de imagens radiográficas pulmonares usando atributos estatísticos e de textura

Abstract

COVID-19 is an infectious disease that became a pandemic in 2020 and one of the methods used to aid in the prognosis were imaging tests, such as chest radiography. One of the major problems of this exam is the difficulty of interpretation, aggravated by the fact that it is a disease still little known and that presents symptoms similar to other pathologies. Considering this, this study aimed to analyze the mean, standard deviation, kurtosis, and skewness and the main Haralick texture features of radiographic images of patients with COVID-19 and of exams reported with different pulmonary diseases and with no radiological finding, in order to contribute to the general characterization of chest radiography exams of this disease and to investigate variables that effectively help in the identification of SARS-CoV-2. As a result of this work, it was possible to notice that COVID-19 exams have statistical and texture features considerably different from those of other pulmonary anomalies, a pattern of local uniformity and homogeneity was observed in chest radiography exams of patients diagnosed with COVID-19, which can be identified by analyzing the highest values of Angular Second Moment, Correlation and Inverse Difference Moment, simultaneously with the lowest values of Contrast and Entropy of Haralick features of these images. In addition, in the COVID-19 exams, a pattern of density and dispersion of pixels was identified, demonstrated by the discrepant higher values of mean and standard deviation. Thus, these variables can be studied with the aim of being used for medical assistance when there is uncertainty in the prognosis during the examination visualization.