Analysis of agreement between radiologists with different levels of CEUS diagnostic experience in CEUSLI-RADS nodule classification: Based on CEUS LI-RADS v2017.

[AIM] This study aimed to assess the reliability of CEUS LI-RADS diagnosis among radiologists with varying levels of CEUS diagnostic experience, and to propose methods to improve the consistency of CEUS LI-RADS classification diagnosis.

[MATERIAL AND METHODS] This retrospective study analyzed CEUS and clinicopathological data of 169 consecutive adult patients (18-89 years old) at high risk of hepatocellular carcinoma (HCC) who were admitted to our hospital between July 2019 to August 2021 and underwent CEUS to analyze the consistency of CEUS LI-RADS diagnoses within and between ultrasound diagnostic observers of different seniority.

[RESULTS] Among the 169 high-risk HCC patients used for consistency analysis, 3, 13, 27, 29, 29, 32, 21, and 15 were LR-1, LR-2, LR-3, LR-4, LR-M, LR-5, TIV, and LR-NC, respectively. After independent assessment, the inter-observer ICC among the three radiologists was 0.797 (95%CI: 0.748-0.840), indicating good reliability. The intra-group ICCs were 0.671 (95%CI: 0.579-0.746) for residents, 0.788 (95%CI: 0.723-0.839) for junior radiologists, and 0.835 (95%CI: 0.783-0.876) for senior radiologists. After unified learning of simple typical CEUS LI-RADS feature images, the ICC between the three radiologists was 0.952 (95%CI:0.938-0.963); the intra-observer ICC for resident was 0.815 (95%CI:0.757-0.860); the intra-observer ICC for junior radiologist was 0.862 (95%CI:0.817-0.896); and the intra-observer ICC for senior radiologist was 0.913 (95%CI:0.884-0.935).

[CONCLUSION] Compared to the independent learning CEUS LI-RADS v2017 guidelines, unified and simplified CEUS LI-RADS typical sign images can effectively improve the reliability of diagnosis among radiologists with different levels of experience.