Value of AI-based Multiomics Analysis in Differentiating COVID-19 from Community-acquired Pneumonia

Objective: To assess the effectiveness of a multiomics model that combines radiomics characteristics and routine clinical information (including clinical symptoms and laboratory data) to distinguish between coronavirus disease 2019 (COVID-19) and community-acquired pneumonia (CAP). Methods: Retrospective data of patients with confirmed COVID-19 caused by the Omicron variant and patients with CAP caused by other viral infections were collected, including chest CT imaging and clinical data. Radiomics, clinical features, and multiomics models were constructed using the entire dataset, and the performance of each model in distinguishing between COVID-19 and CAP was evaluated using receiver operating characteristic curve (ROC) analysis. Results: A total of 8 radiomics features and 7 clinical features were selected to construct the radiomics, clinical features, and multiomics models. The area under the subject operating characteristic curve (AUC) of the radiomics model was 0.759, that of the clinical model was 0.853, and that of the multiomics model was 0.9. Conclusions: The study suggests that AI-based multiomics model has a better performance in differentiating between COVID-19 and CAP compared with those of the radiomics and clinical features models.