Abstract: Objective To evaluate the value of a nomogram based on CT radiomics and clinical risk factors for the preoperative prediction of lymph node metastases of papillary thyroid carcinomas(PTCs)in the central neck. Methods The cases of 114 patients with PTCs diagnosed and treated surgically at Yantai Yuhuangding hospital were retrospectively analyzed, and the clinical and CT imaging data before treatment were collected. The data of the 114 patients were randomly divided into the training(n=85)and test(n=29)sets using a ratio of 7∶3. Radiological features were extracted from the images during the plain CT scan and the enhanced CT arterial phases. Analysis of variance(ANOVA)and the least absolute shrinkage and selection operator(LASSO)algorithm were used to reduce the dimensionality of the radiomics features in the training set to screen out the features with statistical significance. Combining the clinical risk factors and CT-reported lymph nodes status, the efficacy predictors were screened by multivariate logistic regression, and a radiomics nomogram was established for the preoperative prediction of lymph node metastases of PTCs in the central cervical region. The ROC curve was used to evaluate the diagnostic efficiency of the model, and the model was internally verified, calibrated, and clinically applied. Results A total of 2818 CT radiomics features were extracted from the plain and enhanced CT images of 114 patients. After dimensional reduction, 25 features were highly correlated with lymph node metastases of PTCs in the central neck area. The radiomic nomogram, which incorporated the radiomic signature and CT-reported lymph node status, also showed good calibration and discrimination for the test set(AUC 0.858), which were higher than those of the individual CT image nomogram model for the test set(AUC 0.769)those of the prediction model for the individual lymph node status test set(AUC 0.721). The degree of calibration, internal verification consistency, and clinical value were high for this nomogram. Conclusion The presented radiomics nomogram, a non-invasive preoperative tool that incorporates the radiomic signature and CT-reported lymph node status, showed a favorable predictive accuracy for central lymph node metastases in patients with PTCs.

Key words: Papillary thyroid carcinoma, Central lymph node metastasis, Machine learning, Radiomics, Nomogram