THYROID NODULE ASSESSMENT: INTEROBSERVER AGREEMENT OF ULTRASOUND FEATURES AND ACR THYROID IMAGING REPORTING AND DATA SYSTEM SCORES

Abstract

Background: The identification of the nature of the thyroid nodules is mandatory to avoid unnecessary thyroidectomy for the benign nodule and to decrease no of biopsies. TI-RADS scoring helps radiologists in characterization of thyroid nodules and allows simple communication between radiologists, patients, and physicians. Purpose: To assess ultrasound interobserver variability of (ACR) (TIRADS) categories and final scores. Patients and methods: In this study, between June 2019 and July 2022, 553 thyroid nodules in 410 thyroid ultrasound examinations were retrospectively evaluated using the TIRADS features. The definitive diagnosis of benign nodules was based on the fine-needle aspiration or surgical histopathological confirmation. Whereas, the definitive diagnosis of malignant thyroid nodules was based on surgical histopathological confirmation. Additionally, receiver operating curves (ROC) were drawn to compare the actual performance of the observers, and their agreement was tested using Cohen’s kappa. Also, intraclass correlation (ICC) was used to determine the interobserver agreement. Results: Out of 553 thyroid nodules, 452 (81.7%) were benign and 101/553 (18.3%) were malignant. The diagnosis of benign nodules was confirmed by FNA in 324/452 (71.7%) thyroid nodules (Bethesda II) and histopathological analysis in the remaining 128/452 (28.3%) thyroid nodules. Of 101 malignant thyroid nodules that were confirmed by histopathological examination. The majority of benign nodules and TR1 were Bethesda II, whereas Bethesda VI was the most common FNA finding in malignant thyroid nodules and TR5. The interobserver agreement of composition, echogenicity, peripheral calcification, macrocalcification and shape was substantial (k = 0.752, 0.677, 0.727, 0.708, and 0.603 respectively). The agreement of margin, large comet tail, and punctate foci was moderate (k = 0.523, 0.582, 0.484 respectively). There was moderate to almost perfect interobserver agreement for each TIRADS score (K= 0.808, 0.742, 0.591, 0.631, 0.624 for TR1, TR2, TR3, TR4, and TR5 respectively). A substantial agreement was noted for the final TI-RADS score (K= 0.662; 95%CI: 0.614-0.709). Conclusion: ACR TIRADS scoring system has high diagnostic performance. The interobserver agreement of TIRADS categories is moderate to substantial and substantial to almost perfect for various TIRADS scores.