Abstract
RATIONALE AND OBJECTIVES: To develop a CT-based radiomics model to predict central lymph node metastasis (CLNM) in papillary thyroid cancer (PTC) patients and classify risk. MATERIALS AND METHODS: 218 PTC patients from institution 1 were retrospectively enrolled and randomly assigned to a training set and an internal test set (ratio 7:3). Another 64 patients from institution 2 were assigned to an independent test set. Radiomics features were extracted from the arterial phase CT images of PTC. A radiomics signature (Rad-score) was developed using the least absolute shrinkage and selection operator (LASSO) method. Three models, combined model, clinical model, and Rad-score, were established by logistic regression analysis. These models were comprehensively assessed by the area under the receiver operating characteristic curve (AUC), the calibration curve, and the decision curve analysis (DCA). The improvement in predictive efficacy of the combined nomogram was evaluated using the integrated discrimination improvement index (IDI) and net reclassification improvement index (NRI). The defined threshold of the predicted risk score was set at 0.5, and the stratification effect of the combined nomogram was evaluated by subgroup analysis. RESULTS: The Rad-score and another three independent predictors (tumor margin, thyroid capsule state and tumor site) were integrated into a combined nomogram. The AUCs of the combined nomogram were 0.848, 0.858, and 0.840 in the training, internal test, and external test sets, respectively, which were greater than those of the clinical model and the Rad-score. The IDI and NRI were greater than 0 indicating better discriminatory accuracy of the combined nomogram than the clinical nomogram and Rad-score. The net benefit of the combined nomogram in the clinical setting was reflected in the DCA. The combined model allows for the effective stratification of patients in diverse risk subgroups. CONCLUSION: Combining Rad-score and clinical predictors in an integrated model allow for more accurate prediction of CLNM in PTC patients and enables effective risk stratification.