Objective To explore the value of CT-based radiomics and machine learning in predicting the severity of community acquired pneumonia(CAP)in children.Methods The clinical and imaging data of 158 patients diagnosed with CAP in children were analyzed retrospectively.All patients were randomly divided into training set(n=110)and validation set(n=48)in a ratio of 7︰3.Radiomics features were outlined and extracted using 3D Slicer software,and feature selection was achieved using maximum relevance and minimum redundancy(MRMR)and least absolute shrinkage and selection operator(LASSO)algorithms.The construction of the nomogram model and the machine learning combined model was performed by combining clinical features and Radiomics score(Radscore),and its performance was evaluated and validated.Results The area under the curve(AUC)of the clinical model,the radiomics model and the nomogram model in the validation set were classified as 0.82,0.86 and 0.91,respectively.The AUC of the combined multi-layer perceptron(MLP),random forest(RF),and adaptive boosting(ADB)models were 0.926,0.934 and 0.917,respectively in the validation set.Conclusion Radiomics combined with clinical data is expected to be a novel predictor of the severity of CAP in children.MLP,RF and ABD machine learning algorithms can further enable model performance.

Objective To explore the value of CT-based radiomics and machine learning in predicting the severity of community acquired pneumonia(CAP)in children.Methods The clinical and imaging data of 158 patients diagnosed with CAP in children were analyzed retrospectively.All patients were randomly divided into training set(n=110)and validation set(n=48)in a ratio of 7︰3.Radiomics features were outlined and extracted using 3D Slicer software,and feature selection was achieved using maximum relevance and minimum redundancy(MRMR)and least absolute shrinkage and selection operator(LASSO)algorithms.The construction of the nomogram model and the machine learning combined model was performed by combining clinical features and Radiomics score(Radscore),and its performance was evaluated and validated.Results The area under the curve(AUC)of the clinical model,the radiomics model and the nomogram model in the validation set were classified as 0.82,0.86 and 0.91,respectively.The AUC of the combined multi-layer perceptron(MLP),random forest(RF),and adaptive boosting(ADB)models were 0.926,0.934 and 0.917,respectively in the validation set.Conclusion Radiomics combined with clinical data is expected to be a novel predictor of the severity of CAP in children.MLP,RF and ABD machine learning algorithms can further enable model performance.