Artificial Intelligence and radiomics in cervical lymph node staging for head and neck squamous cell carcinoma: a narrative review

Background and objectives. Cervical lymph node (LN) metastasis significantly impacts prognosis in head and neck squamous cell carcinoma (HNSCC), with extranodal extension (ENE) further worsening outcomes. Conventional imaging often lacks sensitivity and specificity, underscoring the need for improved diagnostic tools. Objective. This narrative review evaluates the role of artificial intelligence (AI), including radiomics and deep learning (DL), in enhancing the detection of cervical LN metastasis in HNSCC. Methods. A targeted literature search was conducted to identify recent studies using AI-based models across imaging modalities – contrast-enhanced CT (CECT), diffusion-weighted MRI (DW-MRI), FDG PET/CT, and ultrasound (US). Studies were analyzed for diagnostic performance and clinical utility. Results. AI models demonstrated diagnostic accuracies between 76-92. PET/CT showed improved performance when combined with anatomical imaging. CT-based studies reported AUCs ranging from 0.80 to 0.93, PET-CT studies ranged from 0.95 to 0.99, and one dual-energy CT study reported AUCs between 0.96 and 0.97. These high AUC scores underscore AI’s robust diagnostic potential across modalities. US-based AI, including real-time neural networks, offered accessible, cost-effective options. AI integration improved ENE detection and supported personalized treatment planning. Conclusion. AI techniques, particularly radiomics and DL, show strong potential to enhance cervical LN assessment in HNSCC. Broader clinical adoption requires methodological standardization, external validation, and real-world integration.