Objective:To explore the value of CT radiomics features combined with clinical-radiographic da-ta for differential diagnosis of peripheral small cell lung cancer(SCLC)and non-small cell lung cancer(NSCLC).Methods:The clinical and CT imaging data of peripheral lung cancer patients in our hospi-tal from October 2017 to October 2020 were retrospectively analyzed.All subjects were divided into the SCLC-vs-SCC group and SCLC-vs-ADC group based on the pathology type and then into training and validation sets according to the 7∶3 ratio.The least absolute shrinkage and selection operator re-gression were employed to select radiomics features generated from the training set.The rad-score was calculated for each lesion and the radiomic model was constructed.The clinical-radiographic model was constructed with clinical and imaging features selected by univariate and multivariate re-gression.Based on clinical,CT image,and radiomics features,the comprehensive model was con-structed,and the nomogram was further developed.The diagnostic performances of these models were assessed with the areas under the receiver operating characteristic curve,calibration curve,and decision curve analysis.Results:The comprehensive model indicated good discrimination ability,with AUCs of 0.977(training set)and 0.937(validation set)in the SCLC-vs-SCC group,0.979(training set)and 0.961(validation set)in the SCLC-vs-ADC group,respectively.Decision curve analysis showed that the comprehensive model had higher clinical value than the radiomics and clinical-radiographic models.Decision curve analysis demonstrated good clinical utility.Conclusion:The nomogram combining clinical,CT imaging,and radiomics features provided a potential noninva-sive method to distinguish peripheral SCLC from NSCLC,which could help clinicians make individu-al treatment decisions.
Peripheral Lung CancerSmall Cell Lung CancerTomography,X-Ray ComputedRadiomicsDiagnostic Model
NETL
NSTL
万方数据
