Abstract:

Objective: To evaluate the interobserver variability with the ACR TIRADS（American Radiological Society-Thyroid Imaging Reporting and Data System）and RJ-TIRADS (Ruijin-Thyroid Imaging Reporting and Data System）descriptive category terms of sonographic (US) features ofthyroid nodules. Methods: Retrospective analysis was conducted on 597 nodules. The US findings of each thyroid nodule were analyzed by two radiologists having at least 5 years of experience. All these cases had been pathologically confirmed. Interobserver variability was calculated using Cohen's Kappa statistics. Results: All the descriptive terms had statistical significance. In ACR-TIRADS, there were fair agreement for margin, moderate agreement forecho and calcification; internal structure showed substantial agreement.In RJ-TIRADS, there were no descriptive terms with weak agreement. Moderate agreement was found for echo level and internal blood flow, while substantial agreement was found with regard to margin, boundary, internal structure and calcification. With regard to category, for ACR-TIRADS, TIRADS 1, 2, 3 and 5 showed substantial agreement, while TIRADS 4 showed moderate agreement; for RJ-TIRADS, TI-RADS3, 4C and 5 showed substantial agreement, while TIRADS 4A and 4B showed moderate agreement. If the TIRADS 4 category was evaluated as a whole, the agreement between observerswas better. Conclusions: Both versions of TIRADS have good clinical value in evaluating thyroid nodules. In ACR-TIRADS, the agreement for margin is somewhat weaker, and should be evaluated as a whole or evaluated carefully when used. Both versions of TIRADS require further study to improve the agreement between observers, and classes 4A and 4B in RJ-TIRADS could be combined for elevating the agreement.

Key words: Ultrasound, Thyroid nodule, Thyroid imaging reporting and data system, Interobserver variability