Interrater reliability of MRI Neck Imaging Reporting and Data System (NI-RADS) in the follow-up of nasopharyngeal carcinoma after radiation therapy

Purpose: Evidence supporting the reliability of magnetic resonance imaging (MRI) Neck Imaging Reporting and Data System (NI-RADS) is currently limited. This study aims to evaluate the interrater agreement of MRI NI-RADS among radiologists with varying levels of expertise in nasopharyngeal carcinoma (NPC) patients. Material and methods: We designed an observational retrospective study to identify follow-up MRIs in patients who had undergone radiation therapy. Five radiologists (2 head and neck experts, 1 general radiologist, and 2 residents in radiology) scored each MRI using NI-RADS. Kappa (κ) and percentage of agreement (POA) were calculated for the ultimate score and for each individual feature of the NI-RADS (primary tumor size, signal on T2-weighted images, contrast enhancement, diffusion restriction, and lymph node size). Agreement was analyzed also separately for the first follow-up MRI and subsequent scans. Results: Thirty patients were included (a total of 97 MRIs per rater). Interreader agreement between all readers was moderate for NI-RADS (κ = 0.41; POA = 81%). The first follow-up showed a low reliability between the head and neck expert radiologist and the two radiology residents for both primary tumor contrast enhancement and size assessment (κ = 0.02; POA = 31% and κ = 0.17; POA = 38%, respectively), while there was a high level of agreement in the analysis of diffusion-weighted imaging (DWI) (κ = 0.79; POA = 96%). Conclusion: MRI NI-RADS has a moderate interrater agreement in NPC patients after radiation therapy. Educational effort should focus on the assessment and interpretation of primary tumor contrast enhancement and size in the first examination performed after treatment, by also considering information derived from DWI.