Prediction of PD-L1 expression in non-small cell lung cancer:a dual-phase integrated enhanced CT radiomics model
HAN Xinshu
MA Junli
SHAN Changping
WANG Xun
YANG Jundong
ZHANG Ziqiu
YE Shucheng
Abstract：Objective To construct a dual-phase integrated enhanced CT radiomics model for predicting PD-L1 expression in non-small cell lung cancer(NSCLC)patients.Methods A retrospective study was conducted on 150 NSCLC patients who were pathologically confirmed at the Affiliated Hospital of Jining Medical University from November 2019 to July 2023.These patients were randomly assigned to a training cohort(105 cases)and a testing cohort(45 cases)at a ratio of 7:3.Radiomics features were extracted from both the arterial and venous phases of CT images.Dimensionality reduction and key feature selection were performed using the Least Absolute Shrinkage and Selection Operator(LASSO)algorithm.Eight machine learning algorithms,including logistic regression,were employed to construct radiomics models.The best predictive model was identified through ROC curve analysis,and a dual-phase integrated radiomics model was developed by combining radiomics features from both phases.Univariate and multivariate logistic regression analyses were conducted to evaluate clinical features and to identify independent predictors for constructing a clinical model.A Combine model was then established by integrating radiomics and clinical features.The performance of the models was assessed using ROC curves,and their clinical utility was evaluated using decision curve analysis.Results A total of 1835 radiomics features were extracted from both the arterial and venous phase CT images.After dimensionality reduction and selection,9 radiomics features were ultimately chosen from each phase.Among the radiomics models,the logistic regression model exhibited higher predictive efficiency and robustness.The dual-phase integrated enhanced CT radiomics model demonstrated superior performance compared to single-phase models.The radiomics-clinical model showed the best discriminative ability,with AUC values of 0.822 in the training cohort and 0.681 in the testing cohort.Decision curve analysis indicated the best clinical effectiveness.Conclusion The diagnostic model combining radiomics and clinical features of NSCLC has a good ability to predict PD-L1 expression and can provide a non-invasive and effective diagnostic method for clinical practice.
Keywords：computed tomographynon-small cell lung cancerradiomicsmachine learningPD-L1
Publication Date：2025-09-20
Online Publishing Date：2025-10-22(First online date of this platform, not the publication date of the document)
Pages：10( 1099-1108 )