Back to all papers

A machine learning model based on ultrasound radiomics features combined with clinical features to identify benign thyroid nodules with fibrosis and papillary thyroid carcinoma.

November 21, 2025pubmed logopapers

Authors

Xu M,Xu W,Lu Z

Affiliations (1)

  • Department of Ultrasound, Suzhou Ninth People's Hospital, Suzhou, Jiangsu Province, China.

Abstract

This study investigated whether integrating clinical and radiological ultrasound features of thyroid nodules could improve diagnostic accuracy in distinguishing benign thyroid nodules with fibrosis from papillary thyroid carcinoma. A total of 408 patients who underwent thyroid surgery and had complete ultrasound, clinical, and pathological data were enrolled, including 204 benign nodules with fibrosis and 204 papillary thyroid carcinomas. Regions of interest were manually delineated on ultrasound images using ITK-SNAP, and radiomics features were extracted with Pyradiomics. Patients were randomly divided into a training cohort (n = 326) and a testing cohort (n = 82). Clinical, ultrasound radiomics, and combined nomogram models were constructed, and their diagnostic performance was assessed by sensitivity, specificity, accuracy, and the area under the receiver operating characteristic curve (AUC). In the training cohort, AUCs were 0.860, 0.832, and 0.880 for the clinical, radiomics, and nomogram models, respectively. In the testing cohort, the AUCs were 0.879, 0.814, and 0.926, respectively. Calibration and decision curve analyses demonstrated good consistency and clinical applicability of the nomogram model. The combination of clinical and ultrasound radiomics features significantly improved the diagnostic performance in differentiating benign thyroid nodules with fibrosis from papillary thyroid carcinoma.

Topics

Thyroid Cancer, PapillaryThyroid NoduleMachine LearningThyroid NeoplasmsJournal Article

Ready to Sharpen Your Edge?

Join hundreds of your peers who rely on RadAI Slice. Get the essential weekly briefing that empowers you to navigate the future of radiology.

We respect your privacy. Unsubscribe at any time.