Document Type : Original Article
Department of Medical Informatics, Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran
Nuclear Medicine Research Center, Mashhad University of Medical Sciences ,Mashhad, Iran
Department of Radiology, Faculty of Medicine, Mashhad University of Medical Sciences, Mashhad, Iran
Department of Radiology, Faculty of Medicine, Neyshabur University of Medical Sciences, Neyshabur, Iran
Pharmaceutical Research Center, Mashhad University of Medical Sciences, Mashhad, Iran
Department of Medical Informatics, Amsterdam UMC (location AMC), University of Amsterdam, Amsterdam, The Netherlands
Objective(s): Accurate detection and competent management of thyroid nodules, as a common disease, basically depends on the reliability of the ultrasonography (US) report. In this research, we evaluated inter and intra-observer variation among ultrasonography reporters, based on ACR-TIRADS.
Methods: In this retrospective study, 345 thyroid US images of 150 patients were reviewed. Three clinicians with at least 6-year experience in thyroid US reviewed the images twice at 6-8 weeks’ intervals. Composition, echogenicity, shape, margin, and echogenic foci based on ACR-TIRADS were reported, independently. Inter and intra-observer variations were calculated based on Cohen’s Kappa statistics.
Results: 345 ultrasonography images of 150 patients with thyroid nodules (83 women and 67 men) with a mean age of 65 years were reviewed. Moderate to the substantial intra-observer agreement was achieved with the highest Kapa value in the category of shape (k=0.61-0.77). For TIRADS level, the moderate intra-observer agreement was observed (k=0.42-0.46). Inter-observer agreement for the US category of thyroid nodules was obtained slightly to moderate. Composition
(k=0.42 and 0.51) and echogenicity (k=0.45 and 0.46) showed the highest overall agreement and margin showed the lowest overall agreement (k=0.18 and 0.19). In assessing TIRADS level of nodules, a fair agreement was obtained (k=0.23 and 0.29) .
Conclusion: Moderate to substantial intra-observer agreement and slight to moderate inter-observer variation for evaluation of thyroid nodules; shows the need for a computer-aided diagnosis system based on artificial intelligence to
assist our physicians in differentiating thyroid nodule characteristics based on explicit image features. An additional training course based on ACR-TIRADS for physicians can be another useful recommendation.