Preservation of the facial nerve remains the principal technical challenge in benign parotid surgery. Conventional superficial-versus-deep classification does not predict whether the dominant nerve plane will be encountered over or deep to the tumor. This study evaluated an MRI-based biplanar model anchored to fixed osseous landmarks.

Adults undergoing surgery for benign parotid neoplasms between January 2010 and December 2025 were retrospectively reviewed. Operative findings were the reference standard. Axial deep displacement (ADD), coronal medial offset (CMO), axial posterior proximity, maximum tumor diameter, and cranio-caudal level were assessed. A prespecified logistic model incorporating ADD, CMO, and cranio-caudal level was developed in the 2010–2021 cohort, internally validated with 1,000 bootstrap resamples, and temporally validated in the 2022–2025 cohort.

Of 482 screened cases, 217 were eligible; 124 (57.1%) had an overlying/superior-superficial plane and 93 (42.9%) a deep/inferomedial plane. In a 50-case subset, interobserver agreement was excellent for ADD (ICC 0.941), CMO (ICC 0.965), and cranio-caudal level (weighted κ 0.908). The apparent derivation AUC was 0.849 and the optimism-corrected AUC 0.842. Temporal validation in 52 patients (33 events) yielded an AUC of 0.877 (95% CI 0.777–0.958) and a Brier score of 0.153. Calibration could be estimated only imprecisely (intercept 0.491, 95% CI −0.191 to 1.172; slope 1.173, 95% CI 0.522 to 1.824), and the validation results are therefore preliminary. Zone A predicted an overlying plane in 91.7% of cases, whereas Zone C predicted a deep plane in 81.0%.

A biplanar MRI model based on fixed osseous landmarks provides a reproducible estimate of the dominant facial nerve plane from routine MRI. Discrimination was preserved on temporal validation, whereas calibration remains imprecisely estimated; external validation and prospective evaluation of clinical utility are required before clinical implementation.