Development of a predictive model using key ultrasound characteristics to distinguish follicular thyroid neoplasms from nodular goiter.

Objective: To develop a non-invasive diagnostic model by systematically quantifying and combining key ultrasound imaging characteristics for preoperative distinction. Methodology: We retrospectively analyzed 588 thyroid nodule cases (196 NG, 392 FTN) diagnosed between January 2012 to December 2024 at The First Affiliated Hospital, China. Variables identified through LASSO regression were subjected to multivariate logistic regression, and the resulting predictors were utilized to establish a nomogram. Cross-validation was performed to determine the optimal parameter, specifically the minimum λ (λmin). Model performance was rigorously evaluated using C-index, ROC curves, calibration curves, and decision curve analysis. Ultrasound image acquisition and interpretation were conducted in a blinded manner to minimize potential observer bias. Two-tailed P-values were used, with statistical significance set at P < 0.05. Results: The model demonstrated robust diagnostic performance with an AUC of 0.832 and excellent calibration. The three significant predictors were: single nodule (OR=3.662), presence of peripheral halo (OR=5.338), and rich blood supply (OR=6.075). Decision curve analysis confirmed substantial net benefits across threshold probabilities of 20-89%. Conclusion: This study establishes a clinically practical diagnostic model that effectively integrates and quantifies ultrasound features to differentiate FTN from NG preoperatively. The tool provides a valuable adjunct to conventional assessment, potentially reducing diagnostic errors and preventing unnecessary treatments for thyroid nodules.