Retrospective Evaluation of NI-RADS for Detecting Post-Surgical Recurrence of Oral Squamous Cell Carcinoma on Surveillance CT or MRI.

Background: Imaging surveillance is important for the early diagnosis of recurrence following definitive treatment for oral squamous cell carcinoma (OSCC). The Neck Imaging Reporting and Data System (NI-RADS) provides a standardized template for surveillance imaging and categorizes probability of recurrence at the primary site and the neck (cervical lymph nodes) by assigning categories of 1 (no evidence of recurrence), 2 (low suspicion, subdivided into 2a and 2b for the primary site), 3 (high suspicion), or 4 (definite recurrence). Objectives: The aim of this study was to determine the rate of locoregional and nodal OSCC recurrence stratified by NI-RADS category in patients undergoing surveillance CT or MRI. Methods: This retrospective study included 158 patients enrolled in an institutional surveillance program after resection of OSCC with curative intent. A total of 503 contrast-enhanced CT or MRI examinations performed during surveillance were evaluated. Each examination was randomly assigned to one of four radiologists with expertise in head and neck imaging, who provided NI-RADS categories for the primary site and the neck (1006 assigned NI-RADS categories). NI-RADS performance in identifying recurrence was assessed using receiver operating characteristic (ROC) curve analysis. All four readers evaluated 50 randomly selected cases to determine interreader agreement using Kendall's W. Results: Cancer recurrence was confirmed in 7.6% (38/503) of cases for the primary site and in 6.2% (31/503) for the neck. For the primary site, recurrence rates were 1.0% in NI-RADS category 1, 7.4 % in category 2a, 5.6 % in category 2b, 66.7 % in category 3, and 100.0% in NI-RADS 4. For the neck, recurrence rates were 0.5% in category 1, 7.0% in category 2, 80.0% in category 3, and 100.0% in category 4. NI-RADS demonstrated AUC of 0.934 for the primary site and 0.959 for the neck. Inter-reader agreement was 0.67 for the primary site and 0.81 for the neck. Conclusion: NI-RADS offers excellent discriminatory power in detection of OSCC recurrence, both for the primary site and the neck. Clinical Impact: Radiologists and maxillofacial surgeons should implement NI-RADS in surveillance regimens for post-operative OSCC to help detect recurrences in an effective and standardized fashion by imaging.