ObjectiveTo evaluate the predictive performance of a model integrating ultrasound radiomics features combined with clinical variables for postoperative recurrence in patients with papillary thyroid cancer (PTC) with concomitant Hashimoto thyroiditis (HT). MethodsPatients with PTC with concomitant HT treated at Handan First Hospital from May 2021 to May 2023 were retrospectively enrolled, and the occurrence of postoperative recurrence within two years was assessed. Univariate and multivariate logistic regression models were used to identify risk factors for postoperative recurrence. Ultrasound radiomic features were extracted using the PyRadiomics package. Stable features were selected based on an intraclass correlation coefficient (≥0.75), followed by Z-score normalization on the entire dataset. The Mann-Whitney U test and least absolute shrinkage and selection operator regression were applied to select ultrasound radiomic features, and the radiomics score (Rad-score) was calculated. The clinical variables screened via a multivariate logistic regression model were combined with the Rad-score to develop a nomogram. Internal validation was performed using 1 000 Bootstrap resampling. The predictive performance of the nomogram was quantified using receiver operating characteristic (ROC) curves; calibration performance was verified via calibration curves, and net clinical benefit was determined by decision curve analysis (DCA). ResultsA total of 316 patients were enrolled in this study, among whom 57 experienced recurrence, yielding a recurrence rate of 18.04%. Multivariate logistic regression analysis identified five clinical variables as risk factors for postoperative recurrence in patients with PTC with concurrent HT (all P<0.05): tumor multifocality, maximum nodule diameter > 1 cm, lymph node metastasis, positive thyroglobulin antibody, and thyroid stimulating hormone ≥ 0.38 mU/L. Multicollinearity analysis showed that all variance inflation factors were ＜10, indicating no significant multicollinearity among the independent variables. Through the radiomic feature extraction and selection procedure, 8 key radiomic features with non-zero regression coefficients were selected. Based on these 8 features, the Rad-score was calculated and was significantly higher in patients with recurrence than in those without recurrence [(0.56±0.14) points vs. (0.37±0.09) points, t=12.890, P<0.001]. A nomogram was constructed by incorporating the 5 clinical variables selected by multivariate logistic regression and the Rad-score. The area under the ROC curve of this nomogram was 0.886 [95%CI (0.837, 0.935)]. Internal validation was performed using the Bootstrap method (1 000 resamplings), and the bias-corrected C-index was 0.871. The calibration curve, as assessed by the Hosmer-Lemeshow test (χ2=2.182, P=0.134), demonstrated good agreement between the predicted and actual recurrence probabilities, with a Brier score of 0.088. DCA showed that patients with PTC and HT could benefit from intervention guided by the nomogram within a threshold probability range of 0.21 to 0.84. ConclusionThe nomogram integrating ultrasound radiomic features and clinical variables achieves favorable predictive performance and net clinical benefit for postoperative recurrence in patients with PTC with concomitantT, enabling early postoperative risk stratification and individualized follow-up.