Otolaryngology (ENT)

Multimodal AI predicts post-operative HNSCC outcomes across 1–5 years

In 975 surgically treated HNSCC patients from two centers, an XGBoost model integrated baseline clinicopathology with longitudinal lab data. The model predicted 1–5-year recurrence-free survival with AUCs 0.77–0.83 and sensitivities/specificities generally in the low 70% range. Overall survival prediction was similar, with AUCs around 0.79–0.82 across timepoints and strong performance in HPV-positive oropharyngeal cancer. Non-HPV HNSCC maintained OS AUCs 0.78–0.81 and RFS AUCs 0.77–0.83, suggesting broad applicability across subsites. These results support AI-based dynamic risk scoring as an adjunct for surveillance intensity and adjuvant treatment discussions after surgery. 1

Multimodal deep learning improves staging accuracy in LSCC

This multicenter study developed a deep learning model combining clinical data, contrast-enhanced CT, and laryngoscopy to separate early (I–II) from advanced (III–IV) LSCC. Among 450 pathologically confirmed cases, the integrated model achieved AUCs of 0.90 (internal) and 0.89 (external), outperforming all single-modality approaches. Calibration and decision curve analyses suggested better reliability and net clinical benefit than CT-only, laryngoscopy-only, or clinical models. Model-derived risk groups showed significantly different progression-free survival, indicating added prognostic value for treatment planning. If validated further, this tool could help tailor laryngeal preservation strategies and radiotherapy planning in borderline cases. 2

Population-based data define incidence and outcomes in parotid carcinoma

Using the Swedish Head and Neck Cancer Register, this study analyzed 1,018 parotid carcinomas diagnosed from 2008–2019. Age-adjusted incidence remained stable at about 0.9 cases per 100,000 person-years, with 90% receiving curative-intent treatment. Overall three-year recurrence was 9%, highest in salivary duct carcinoma and adenocarcinoma, and lower in acinic cell and mucoepidermoid carcinoma. For stage I–II tumors, five-year survival was unaffected by whether malignancy was known preoperatively or discovered postoperatively. Male sex, older age, stage III–IV disease, and ECOG 2–4 increased overall mortality, while timing of adjuvant radiotherapy did not. 3