Model based on the automated AI-Driven CT quantification is effective for the diagnosis of refractory Mycoplasma pneumoniae pneumonia

Abstract

Refractory Mycoplasma pneumoniae pneumonia（RMPP）prediction is a challenging but clinically significant challenge. A model based on AI-derived quantitative determination of lung lesions extent on initial computed tomography (CT) scan and clinical indicators has the potential to facilitate early RMPP prediction in hospitalized children. In this study, we conducted a retrospective cohort as a training set including 126 children with M. pneumoniae pneumonia (MP) admitted to Children’s Hospital of Nanjing Medical University, China from January 2019 to December 2020. We defined a constant Φ which can combine the volume and CT value of pulmonary lesions and be further used to calculate the logarithm of Φ to the base of 2 (Log2Φ). Finally, a clinical-imaging prediction model was developed using Log2Φ and clinical characteristics. The area under the receiver operating characteristic curve (ROC-AUC) was applied to performance evaluation. Then we conducted a prospective study including 54 children with MP as a test set to validate the predictive model for RMPP from January to December 2021. The clinical model yielded the AUC of 0.810 and 0.782, and the imaging model yielded the AUC of 0.764 and 0.769 in the training set and the test set, respectively. The clinical-imaging model combining Log2Φ, temperature(T), aspartate aminotransferase (AST), preadmission fever duration (PFD), and preadmission macrolides therapy duration (PMTD) produced the highest AUC values of 0.897 and 0.895 in the training set and the test set, respectively. Our work demonstrated that using automated quantification of lung disease at CT combined with clinical data in MPP is useful to predict RMPP.