Establishment of a multiparametric PET/CT-based diagnostic model for lymph node metastasis in lung cancer and its clinical value
QIAO Tingting
GAO Lin
CUI Xiao
ZHAO Xiaowen
CHENG Wenyue
LI Kun
CHENG Zhaoping
DUAN Yanhua
Abstract：Objective To construct and evaluate the diagnostic performance of multiparametric PET/CT-based predictive model in predicting lymph node metastasis(LNM)of lung cancer.Methods A total of 106 lung cancer patients who underwent fluorodeoxyglucose(¹⁸F-FDG)PET/CT were enrolled in the retrospective study.Based on LNM status,they were divided into the positive group(n=68)and the negative group(n=38).A total of 500 observable lymph nodes with increased FDG uptake were in-cluded,with 362 from the positive group(training/test cohort:254/108)and 138 from the negative group(training/test cohort:96/42).Binary logistic regression and LASSO regression analyses were utilized to identify the independent predictive factors and to construct the predictive models.The predictive performance was evaluated using the area under the receiver operating charac-teristic curve.Results Logistic regression analysis results showed that SUVmax(P<0.001)and CT density(P<0.001)were confirmed as independent predictors of LNM.In the training cohort,the AUC of Standard CT-Based model,PET model,and multiparametric PET/CT model were 0.816,0.882,and 0.914,respectively.In the test cohort,the AUC was 0.790,0.880,and 0.918,respectively.The Hosmer-Lemeshow test confirmed that the multiparametric predictive model demonstrated an excellent fit(R²=0.575;χ²=86.39).Clinical decision curve analysis(DCA)for both the training and test cohorts indicated that multipara-metric PET/CT model provided a better clinical net benefit than other models.Conclusions Multiparametric PET/CT model demonstrates excellent diagnostic performance for LNM in lung cancer patients.
Keywords：Lung cancerPositron Emission Tomography/Computed TomographyLymph node metastasisPredictive model
Publication Date：2025-03-30
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：5( 43-47 )