Artificial intelligence-based chest X-ray (AI-CXR) and hematology parameters to predict mortality and intubation events In COVID-19 patients at the second-referral hospital in Indonesia

Abstract

COVID-19 can cause acute respiratory distress syndrome (ARDS), and artificial intelligence (AI) algorithms for chest X-ray (CXR) analysis have been developed to assess pulmonary damage in COVID-19 patients. We evaluated the combination of AI-CXR and hematological parameters to predict mortality and intubation events for COVID-19 patients. This study is a retrospective cohort study of COVID-19 patients for which we have collected data during the 2020–2022 period at Airlangga University Hospital, Surabaya, Indonesia. The total number of patients involved in this study was 312. The results of the scoring evaluation using a combination of AI-CXR, hematology parameters, respiratory rate (RR), and SaO2 showed ROC 0.854; p = 0.000; sensitivity: 74.2%; specificity: 89% in predicting 30-day mortality events, and ROC 0.846; p = 0.000; sensitivity: 80.3%; specificity: 80.9% in predicting 30-day intubation events. Meanwhile, without using AI-CXR, it shows ROC 0.822; p = 0.000; sensitivity: 80.3%; specificity: 81.7% in predicting 30-day mortality events, and ROC 0.815; p = 0.000; sensitivity: 77.3%; specificity: 81.7% in predicting 30-day intubation events. The combination of hematology parameters with or without AI-CXR has excellent sensitivity and specificity in predicting the incidence of mortality and intubation in COVID-19 patients. Not applicable.

Ähnliche Arbeiten