The Diagnostic Value of Artificial Intelligence in C-TIRADS 4-5 Nodules, Real-Time Dynamic Ultrasound and Contrast-Enhanced Ultrasound to Enhance the Difference Between Papillary Thyroid Carcinoma and Nodular Goiter.

BACKGROUND

Differentiating papillary thyroid carcinoma (PTC) from nodular goiter (NG) in thyroid nodules is challenging. Advanced tools such as contrast-enhanced ultrasound (CEUS) and artificial intelligence (AI)-assisted diagnostics may improve diagnostic accuracy for Chinese Thyroid Imaging Reporting and Data System (C-TIRADS) 4-5 nodules.

OBJECTIVE

To evaluate the diagnostic performance of conventional ultrasound (CUS), CEUS, and AI dynamic ultrasound in distinguishing PTC from NG in C-TIRADS 4-5 nodules.

METHODS

This retrospective, single-center study included 180 PTC and 158 NG patients. Diagnostic performance was assessed using the area under the receiver operating characteristic curve (AUC), with statistical comparisons conducted via bootstrapping methods (1000 iterations) implemented in Python 3.12.6.

RESULTS

The individual models demonstrated strong diagnostic performance, with AUCs of 0.85 (C-TIRADS), 0.86 (CEUS), and 0.86 (dynamic AI). Combining models enhanced sensitivity but reduced specificity. The majority voting system, incorporating all three models, achieved the highest diagnostic performance (AUC 0.93, sensitivity 97%, specificity 89%, accuracy 93%). No significant differences were observed between AUCs due to the strong discriminatory ability of each method.

CONCLUSION

All models, including C-TIRADS, CEUS, and dynamic AI, performed well in differentiating PTC from NG. Combining these methods, particularly with majority voting, improved diagnostic performance without compromising specificity.