Comparison of Chest Computed Tomography-based Severity Scoring Systems for Detecting Lung Involvement in Patients with Coronavirus Disease 2019

Objective: This study aimed to evaluate the burden of lung involvement in patients with coronavirus disease 2019 (COVID-19) infection using four semi-quantitative measurement methods; compare the diagnostic performance of the four methods and consistency of results over time among observers; and find the fastest and most accurate evaluation method. Methods: Data of 157 patients with COVID-19 infection confirmed by real-time reverse transcription polymerase chain reaction (RT-PCR) using nasopharyngeal swab samples and hospitalized were retrospectively analyzed. According to the "Diagnosis and Treatment Plan for COVID-19 Infection (Tenth Edition)" issued by the China's National Health Commission, 87 patients were classified as medium type, 66 as severe, and 4 as critical. Based on the patients' chest CT images, two radiologists independently evaluated the severity of lung involvement using four semi-quantitative scoring systems and recorded the time taken by each scoring system. The intra-group correlation coefficient (ICC) was used to test the evaluation results consistency between the two radiologists using four semi-quantitative evaluation methods. Result: There were statistically significant differences in D-dimer, lymphocyte percentage, lymphocyte count, lactate dehydrogenase, and C-reactive protein between the severe and non-severe groups. The scoring results of the four semi-quantitative evaluation methods showed that the average score in the severe group was higher than that in the non-severe one. The consistency of the four semi-quantitative evaluation methods used by the two radiologists was poor, of which the ICC of the four methods was less than 0.75. The total severity score (T-SS) method showed the shortest average time. Conclusion: The semi-quantitative assessment method for detecting severity of pulmonary lesions in patients infected with COVID-19 based on vision and subjective experience has some limitations. The development of a quantitative assessment model based on artificial intelligence with good generalization in the future has important clinical application value.