Comparison of chest radiological findings between coronavirus disease 2019 pneumonia and anti-melanoma differentiation-associated gene 5 antibody- positive interstitial pneumonia by artificial intelligence-based quantitative computed tomography image analysis

Background: Coronavirus disease 2019 (COVID-19) pneumonia and anti-melanoma differentiation-associated gene 5 antibody-positive interstitial pneumonia (MDA5-IP) share many similarities; however, the treatment and management of the two diseases are different. In the early stages of developing a treatment plan, it is crucial to distinguish between the two diseases. This study was conducted to compare the radiological findings between COVID-19 pneumonia and MDA5-IP. Methods We recruited patients with COVID-19 pneumonia between January and June 2020. The control group comprised patients with MDA5-IP admitted between April 2013 and December 2019. Patients with thin-slice computed tomography (CT) images within 2 days of admission were enrolled. The CT images were analyzed using an artificial intelligence-based quantitative CT software program. Radiological findings were classified as faint ground-glass opacity (GGO), GGO, reticulation, consolidation, honeycombing, nodules, hyperlucency, or interlobular septum. The volumes of these radiological findings were compared between the two groups. A classification and regression tree algorithm was used to develop a prediction model to stratify the risk of COVID-19 pneumonia. Results We enrolled 72 and 15 patients in the COVID-19 pneumonia and MDA5-IP group, respectively. Faint GGO and consolidations were observed more extensively in patients with MDA5-IP. The prediction model was developed at cut-off values of faint GGO, < 30%; GGO, ≥ 10%, and consolidation < 1%. This prediction model contributed to changing post-test probability in 26% of cases. Conclusion The COVID-19 group showed fewer faint GGO and consolidation volumes than the MDA5-IP group. We developed a predictive model to stratify the risk of COVID-19 pneumonia.