Imaging Study of COVID-19 Patients with Diabetes Mellitus by Computed Tomograpgh Quantitative Indicators Based on Deep Learning

Objective: To investigate the imaging characteristics of coronavirus disease 2019 (COVID-19) patients with diabetes mellitus using deep learning-based quantitative computed tomograpgh (CT) indicators. Materials and methods: The clinical and imaging data of 112 COVID-19 patients admitted to the Department of Infection, Beijing Shijitan Hospital, Capital Medical University, from December 2022 to January 2023 were retrospectively collected. The patients were divided into diabetic and non-diabetic groups according to their diabetes history, and the clinical and quantitative CT imaging characteristics of the two groups were analyzed using univariate analysis. Results: A total of 112 patients with COVID-19, aged 26-95 years (mean, (70.4±14.4) years), were included in the study, and 44.6% (50/112 cases) were female. In terms of clinical features, C-reactive protein levels were significantly higher in the diabetic group. In terms of CT quantitative indicators, patients in the diabetic group had higher number of whole lung and left lung lesions, lesion volume, and mediastinal lymph nodes than patients in the non-diabetic group. In addition, patients in the diabetic group had a larger volume of ground glass opacity and solid opacity, and patients in the diabetic group had a smaller volume ratio of ground glass opacity and solid opacity in terms of imaging signs, patients in the diabetic group had a higher proportion of lesions with large patchy and banded patterns, and they had a higher proportion of halo signs, air bronchial signs, air trapping signs, mosaic perfusion signs, and subpleural black bands. Conclusion: Pulmonary lesions in patients with diabetes combined with COVID-19 have characteristic features, and deep learning-based quantitative CT indicators, particularly the solid opacity observed in the lungs, can provide valuable information on the extent and severity of lesions in these patients.