Abstract:

Objective To construct a combined model integrating ultrasonic features and radiomics derived from ultrasound images， and to evaluate its diagnostic performance in re - assessing the benign or malignant nature of Bethesda Ⅲ nodules. Methods A retrospective study was carried out on 442 patients with thyroid nodules classified as Bethesda Ⅲ after fine-needle aspiration biopsy （FNAB） between January 2019 and September 2024. All patients had undergone surgical pathology. The patients were randomly allocated into a training set and a testing set at a ratio of 7∶3. Relevant clinical characteristics were gathered， and regions of interest （ROI） were outlined on the most suspicious slice of the lesion prior to biopsy. Ultrasound radiomics features were extracted， key radiomics features were selected， and radiomics scores （Rad-score） were computed. The ultrasound model， radiomics model， and combined model were constructed. Subsequently， the diagnostic efficacy and clinical application value of each model were evaluated using the area under the receiver operating characteristic curve （AUC） and decision curve analysis （DCA）. Results Univariate analysis and multivariate logistic regression analysis findings indicated that microcalcification， irregular margin， and Rad-score were independent risk factors for the malignant transformation of Bethesda Ⅲ nodules. In the testing set， the AUC values of the ultrasound model， radiomics model， and combined model were 0.76， 0.71， and 0.81， respectively. The calibration curve of the combined model revealed a good consistency between the predicted values and the actual outcomes. The DCA of the testing set demonstrated that the combined model exhibited high clinical utility. Conclusion The combined model， established based on ultrasonic features and ultrasound radiomics， provides a higher predictive value for evaluating the malignancy risk of Bethesda Ⅲ nodules.

Key words: Bethesda Ⅲ nodules, radiomics, thyroid, ultrasound