Is histogram analysis useful in the diagnosis of COVID-19 patients?

Abstract

Aim: In this study, we aimed to show the contribution of the chest computed tomography (CT)-based histogram analysis method, which will enable us to make quick decisions for patients who are clinically suspected of having COVID-19 infection and whose diagnoses cannot be confirmed by polymerase chain reaction (PCR) tests. Material and Methods: A total of 84 patients, 40 in the PCR-positive group (age range: 17-90 years) and 44 in the PCR-negative group (age range: 15-75 years), were included in the study. A total of 154 lesions with ground-glass density, 78 in the PCR-positive group and 76 in the PCR-negative group, were detected in these patients' thorax CT scans. The region of interest was placed on the ground-glass opacities from the images and numerical data were obtained by histogram analysis. Numerical data were uploaded to the MATLAB program. Results: The localizations of ground-glass densities in the CT findings of patients with probable and definite COVID-19 diagnoses were similar; 74.7% of the ground-glass areas in both groups showed peripheral distribution. Lesions were frequently observed in right lungs and lower lobes. In histogram analysis, standard deviation, variance, size %L, size %M, and kurtosis values were higher in the PCR-positive than the PCR-negative group. When receiver operating characteristic curve analysis was performed for standard deviation values, the area under the curve was 0.640, and when the threshold value was selected as 123.4821, the two groups could be differentiated with 62.8% sensitivity and 61.8% specificity. Discussion: The use of histogram-based tissue analysis, which is a subdivision of artificial intelligence, for clinically highly suspicious patients increases the diagnostic accuracy of CT. Therefore, performing CT analysis with the histogram method will significantly aid healthcare professionals, especially in clinics where rapid decisions are required, such as in emergency services.