A comprehensive predictive model study: ultrasound viscosity imaging in thyroid nodules

Abstract Purpose This study aimed to develop a comprehensive predictive model that integrates clinical factors, conventional ultrasound (US) features, and ultrasound viscosity imaging metrics to improve diagnostic accuracy in thyroid nodules and to explore the role of viscosity metrics in differentiating benign and malignant thyroid nodules. Methods A total of 215 patients with thyroid nodules were included in the study, with 70% in the training set and 30% in the testing set. Univariate and multivariate logistic regression were employed to identify key predictors, resulting in the development of three models. Three models were developed: ModA (clinical factors), ModB (clinical factors and conventional US features), and ModC (clinical factors, conventional US features, and US viscosity imaging). The diagnostic efficacy of the models was evaluated and compared using receiver operating characteristic curves. Results Multivariate analysis revealed that factors, such as echogenicity, viscosity minimum (Vmin), and viscosity mean (Vmean), were significant risk factors for distinguishing between benign and malignant thyroid nodules. In the testing set, ModA demonstrated an Area Under the Curve (AUC) of 0.777, ModB exhibited an improvement to 0.922, and ModC demonstrated an AUC of 0.945. The findings indicate that ModC exhibits superior predictive accuracy for differentiating between benign and malignant thyroid nodules. Conclusion The integration of US viscosity imaging with clinical factors and conventional US features can enhance the predictive performance of differentiation in thyroid nodules, offering a reliable clinical tool to support early and accurate diagnosis of thyroid nodules. Research Questions