[Study on the predictive value of a multimodal ultrasound feature fusion model for central lymph node metastasis of...

Objective:To investigate the predictive value of a multimodal ultrasound feature model combined with clinical indicators for the preoperative assessment of central lymph node metastasis（CLNM） in patients with papillary thyroid microcarcinoma（PTMC）. Methods:The clinical and preoperative multimodal ultrasound data of 120 PTMC patients from February 2025 to July 2025 treated in our hospital were retrospectively collected, including two-dimensional ultrasonography（2D-US）, microvascular imaging（MVI）, and shear wave elastography（SWE）. According to postoperative pathological results, patients were divided into a metastasis group（45 cases） and a non-metastasis group（75 cases）. The clinical and ultrasound characteristics between the two groups were compared. Variables with statistically significant differences were further evaluated for multicollinearity and subsequently included in multivariate logistic regression analysis to identify independent risk factors for CLNM and construct predictive models. Receiver operating characteristic（ROC） curves were plotted to evaluate the predictive performance of the models, and their diagnostic efficacy was compared by DeLong test. Results:The results of univariate analysis showed that age, gender, TSH level, maximum diameter of the tumor, multifocality, tumor location, capsule contact, boundary, shape, microcalcification characteristics, specific vascular patterns in MVI, and SWE parameters（E, Emin, V, and Vmin） were associated with CLNM（P<0.05）. The consequences of multivariate logistic regression analysis indicated that TSH level, tumor multifocality, contacting the capsule, and the minimum shear wave velocity（Vmin） were independent risk factors for CLNM in PTMC（OR>1, P<0.05）. The multimodal ultrasound fusion model demonstrated the optimal comprehensive predictive efficacy （AUC=0.880）, which was significantly superior to the 2D-US model（AUC=0.808）, the SWE model（AUC=0.676）, and the MVI model（AUC=0.591）, and exhibited the best balance of sensitivity and specificity. Conclusion:TSH level, multifocality, capsule contact, and Vmin are important predictors of CLNM in PTMC. A multimodal ultrasound-based predictive model can significantly improve the preoperative prediction of CLNM and may provide valuable imaging evidence for risk stratification and clinical decision-making in patients with PTMC.