Radiological-pathological correlation of Head and Neck Cancer International Group (HNCIG) criteria for extranodal extension across imaging modalities: A retrospective cohort study

Objectives The Head and Neck Cancer International Group (HNCIG) criteria standardized the evaluation of imaging-detected extranodal extension (iENE) in head and neck squamous cell carcinoma (HNSCC). We aimed to validate the radiologic–pathologic correlation of HNCIG criteria compared with non-standardized (NS) evaluation. Methods We retrospectively included 176 patients who underwent neck dissection for HNSCC between August 2019 and October 2024. Two experienced radiologists (ERs) and two trainees (TRs), blinded to pathology, independently reviewed pre-operative computed tomography (CT) and/or magnetic resonance imaging (MRI) using NS and HNCIG criteria (G0-G3). Diagnostic performance was evaluated using receiver operating characteristic analysis with area under the curve (AUC). Paired AUC comparisons were performed using DeLong’s test with bootstrap resampling. Interobserver reliability was assessed using kappa statistics. Results We evaluated 67 CT (22 pENE + ) and 176 MRI (32 pENE + ). HNCIG ≥ 1 criteria showed excellent diagnostic performance with AUCs of 0.97 (95% CI, 0.93–0.99) for MRI-ERs, 0.92 (95% CI, 0.84–0.99) for CT-ERs, 0.92 (95% CI, 0.86–0.96) for MRI-TRs, and 0.94 (95% CI, 0.89–0.98) for CT-TRs. Compared with NS evaluation, HNCIG achieved significantly higher AUCs on MRI for ERs (p = 0.002) and TRs (p = 0.019) and on CT for TRs (p < 0.001). Interobserver agreement was moderate on CT (HNCIG κ = 0.60; NS κ = 0.62) and substantial on MRI (HNCIG κ = 0.75; NS κ = 0.74). Conclusion Despite the monocentric retrospective design, HNCIG criteria outperformed NS evaluation on MRI regardless of reader experience and on CT for trainees.