Preoperative ultrasound risk stratification of submandibular gland lesions: a retrospective cohort study

Background Submandibular gland lesions are rare and diagnostically challenging because benign neoplasms, inflammatory mass lesions, carcinomas, and lymphomas may present with overlapping sonographic features. While ultrasound is central to the preoperative work-up, most malignancy prediction models were developed in mixed major salivary gland cohorts dominated by parotid tumors. Their performance in the submandibular gland remains unclear. Methods We conducted a retrospective cohort study of consecutive patients who underwent submandibulectomy for an intraglandular lesion at a tertiary referral center between 2019 and 2025. All patients had standardized preoperative ultrasound and histopathologic confirmation. Ultrasound images were reviewed while blinded to clinical and histopathological outcomes. Clinical variables, conventional sonographic features, the taller-than-wide (TTW) ratio, the score proposed by Lo et al., and a TTW-modified version of this score were evaluated for their ability to discriminate benign from malignant lesions. Results A total of 120 patients were included (mean age 58.4 ± 17.4 years; 75% male). Final histology showed 96 benign lesions (80.0%), 13 carcinomas (10.8%), and 11 lymphomas (9.2%). Among clinical features, only nerve palsy was associated with malignancy (4.2% vs. 0%, p = 0.04). On ultrasound, malignant lesions more frequently showed irregular margins (37.5% vs. 17.7%, p = 0.04) and suspicious cervical lymph nodes (20.8% vs. 1.0%, p = 0.001). Other ultrasound features were not significantly different between benign and malignant lesions. The TTW ratio differed in exploratory subgroup analysis across benign tumors, carcinomas, and lymphomas ( p = 0.02), but not in the overall benign-versus-malignant comparison. The Lo et al. score showed poor discrimination (AUC 0.62, 95%-CI 0.48–0.76), and the TTW-modified score performed similarly (AUC 0.61, 95%-CI 0.47–0.74). Conclusion Preoperative ultrasound is valuable in the diagnostic assessment of submandibular gland lesions, particularly for identifying irregular margins and suspicious cervical lymph nodes, but morphology alone has limited ability to distinguish benign from malignant disease. Parotid-derived ultrasound risk scores, including TTW-based modification, showed poor transferability to the submandibular gland.