Comparative Analysis of ACR TI-RADS and K-TIRADS: Inter-System Agreement and Diagnostic Performance Using a Single Study Cohort

Abstract

Purpose To evaluate the inter-system agreement and diagnostic performance of experienced readers using two Thyroid Imaging Reporting and Data Systems (TI-RADS). Materials and Methods A total of 481 thyroid US images were collected retrospectively. Four experienced radiologists independently reviewed the images and documented the US features and final assessments based on the Korean TI-RADS (K-TIRADS). The final American College of Radiology (ACR) TI-RADS score was calculated using individual US feature scores. The final assessments from both TI-RADS systems were used to evaluate inter-reader and inter-system agreement. The diagnostic performance metrics, including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and unnecessary biopsy rate, were calculated and compared. Results Among the 481 US images, 157 were benign, 184 were malignant, and 140 were negative images. In the final assessment, substantial agreement was observed between the two TI-RADS systems (kappa = 0.61) for all four readers. The mean AUC did not differ significantly between ACR TI-RADS and K-TIRADS (p = 0.52). Compared with K-TIRADS, ACR TI-RADS demonstrated significantly higher sensitivity and NPV but significantly lower specificity, PPV, and unnecessary biopsy rates (p < 0.001). Conclusion Substantial agreement was observed between the final assessments of ACR TI-RADS and K-TIRADS. The mean AUC did not differ significantly between the two TI-RADS categories, reflecting consistent image interpretation and patient management regardless of the system.