Application value of a Nomogram model established on clinical and ultrasound features in modifying classification results of the Chinese Thyroid Imaging Reporting and Data System by ultrasound physicians

Objective To analyze the application value of a Nomogram model established on clinical and ultrasound features of thyroid in modifying classification results of the Chinese Thyroid Imaging Reporting and Data System (C-TI-RADS) by ultrasound physicians.

Methods The clinicopathological ultrasound materials of patients with surgical resection for thyroid nodules in Sichuan Provincial People's Hospital (training set, n=841) and Sichuan Mianyang 404 Hospital (external validation set, n=295) from January 2021 to December 2022 were retrospectively analyzed, and preoperative thyroid ultrasound was used to classify thyroid nodules by C-TI-RADS. Univariate and Multivariate Logistic regression analyses were used to screen independent predictors in the training set and a Nomogram model was constructed, and Bootstrap resampling was used for internal validation; ultrasound physicians in Sichuan Mianyang 404 Hospital were charged in external verification based on this Nomogram model. Receiver operating characteristic (ROC) curve and calibration curve were drawn to evaluate the efficacy of the model and its clinical value in modifying the classification results by C-TI-RADS.

Results Univariate and Multivariate Logistic regression analyses showed that gender, age, maximum diameter of nodules, the number of nodules, ultrasound abnormalities of cervical lymph nodes and C-TI-RADS classification were the independent factors for predicting the modified classification results of C-TI-RADS by ultrasound physicians (P < 0.05). Consistency index of the Nomogram model constructed based on these factors was 0.842 (95%CI, 0.816 to 0.867), the area under the curve (AUC) was 0.842, sensitivity was 92.9%, specificity was 63.7%, and accuracy was 75.9%. The prediction results of the Nomogram model were in good agreement with the actual situation in both the training set and the external validation set.

Conclusion Nomogram model based on clinical and ultrasound features of thyroid shows good predictive accuracy in modifying classification results of C-TI-RADS by ultrasound physicians, which has potential clinical application value.