Computed tomography multi-phase radiomics models for differentiating benign and malignant thyroid nodules
4 affiliations
- 1Jinhua Graduate Joint Training Base, Zhejiang Chinese Medical University
- 2Department of Radiology, Wenzhou Medical University Affiliated Quzhou Hospital (Quzhou People’s Hospital)
- 3Department of Quzhou Key Laboratory of Respiratory Mechanics in Critical Care Medicine, Wenzhou Medical University Affiliated Quzhou Hospital (Quzhou People’s Hospital)
- 4Department of Radiology, Ningbo No. 2 Hospital
Abstract
Objective: To develop and validate an integrated model combining multiphase CT radiomics features, clinical data, and conventional morphological CT signs for the preoperative differentiation of benign and malignant thyroid nodules.
Methods: We retrospectively enrolled patients with surgically confirmed thyroid nodules who underwent preoperative contrast-enhanced thyroid CT examination at two centers. Patients from center 1 were randomly allocated to a training cohort and an internal validation cohort at an 8:2 ratio; patients from center 2 served as an external validation cohort. Clinical characteristics and morphological features of nodules were assessed. Radiomics features were extracted from all three CT phases, and four categories of models were constructed: clinical model, single-phase CT radiomics models, multiphase CT radiomics model, and combined clinical-radiomics model. Model performance and interpretability were evaluated using receiver operating characteristic (ROC) curve analysis, calibration curves, decision curve analysis (DCA), and SHapley Additive exPlanations (SHAP).
Results: The multiphase CT radiomics model yielded areas under the ROC curve (AUCs) of 0.919, 0.836, and 0.787 in the training, internal validation, and external validation cohorts, respectively. Among the single-phase models, the venous-phase model achieved the best external validation performance (AUC = 0.815). The combined model, integrating clinical and morphological features with multiphase radiomics features, achieved the highest overall diagnostic performance, with an external validation AUC of 0.865 (sensitivity = 74.0%, specificity = 84.5%), significantly outperforming both the clinical model (AUC = 0.800) and the multiphase radiomics model (AUC = 0.787). It also demonstrated favorable calibration and greater clinical net benefit.
Conclusion: Multiphase CT radiomics enables effective differentiation of benign and malignant thyroid nodules. Integration of clinical and morphological information further improves diagnostic performance, providing an objective, noninvasive tool for preoperative risk stratification of incidentally detected thyroid nodules on CT.
Related
- Differentiating pancreatic acinar cell carcinoma from solid pseudopapillary neoplasm using contrast-enhanced computed tomography
- Ultrasound-based deep learning radiomics with explainable machine learning for predicting benign and malignant thyroid nodules in lymphocytic thyroiditis: a multicenter study
- Nomogram models integrating multiregion radiomics, quantitative computed tomography, and clinical indicators for distinguishing connective tissue disease-associated interstitial lung disease from idiopathic pulmonary fibrosis and improving the prediction of gender-age-physiology stage
- Whole-phase DCE-MRI radiomics for differentiating benign and malignant BI-RADS 4 breast lesions: a comparative machine learning study
- Differentiating Chlamydia psittaci pneumonia from community-acquired pneumonia using clinical features and computed tomography findings
BibTeX
@article{li2026computed,
title = {{Computed tomography multi-phase radiomics models for differentiating benign and malignant thyroid nodules}},
author = {Ning Li and Jianan Tong and Guozheng Zhang and Shufeng Xu and Weitao Huang and Xiaowei Han and Jingfeng Zhang and Xisong Zhu},
journal = {Frontiers in Endocrinology},
year = {2026},
month = oct,
doi = {10.3389/fendo.2026.1928263},
url = {https://www.frontiersin.org/articles/10.3389/fendo.2026.1928263/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Computed tomography multi-phase radiomics models for differentiating benign and malignant thyroid nodules
AU - Ning Li
AU - Jianan Tong
AU - Guozheng Zhang
AU - Shufeng Xu
AU - Weitao Huang
AU - Xiaowei Han
AU - Jingfeng Zhang
AU - Xisong Zhu
PY - 2026
DA - 2026/10/15
JO - Frontiers in Endocrinology
DO - 10.3389/fendo.2026.1928263
UR - https://www.frontiersin.org/articles/10.3389/fendo.2026.1928263/full
AB - Objective: To develop and validate an integrated model combining multiphase CT radiomics features, clinical data, and conventional morphological CT signs for the preoperative differentiation of benign and malignant thyroid nodules. Methods: We retrospectively enrolled patients with surgically confirmed thyroid nodules who underwent preoperative contrast-enhanced thyroid CT examination at two centers. Patients from center 1 were randomly allocated to a training cohort and an internal validation cohort at an 8:2 ratio; patients from center 2 served as an external validation cohort. Clinical characteristics and morphological features of nodules were assessed. Radiomics features were extracted from all three CT phases, and four categories of models were constructed: clinical model, single-phase CT radiomics models, multiphase CT radiomics model, and combined clinical-radiomics model. Model performance and interpretability were evaluated using receiver operating characteristic (ROC) curve analysis, calibration curves, decision curve analysis (DCA), and SHapley Additive exPlanations (SHAP). Results: The multiphase CT radiomics model yielded areas under the ROC curve (AUCs) of 0.919, 0.836, and 0.787 in the training, internal validation, and external validation cohorts, respectively. Among the single-phase models, the venous-phase model achieved the best external validation performance (AUC = 0.815). The combined model, integrating clinical and morphological features with multiphase radiomics features, achieved the highest overall diagnostic performance, with an external validation AUC of 0.865 (sensitivity = 74.0%, specificity = 84.5%), significantly outperforming both the clinical model (AUC = 0.800) and the multiphase radiomics model (AUC = 0.787). It also demonstrated favorable calibration and greater clinical net benefit. Conclusion: Multiphase CT radiomics enables effective differentiation of benign and malignant thyroid nodules. Integration of clinical and morphological information further improves diagnostic performance, providing an objective, noninvasive tool for preoperative risk stratification of incidentally detected thyroid nodules on CT.
SN - 1664-2392
ER -