AI techniques show promise for staging head and neck cancer

This scoping review examined 44 studies published between January 2015 and January 2026. The research looked at how artificial intelligence techniques, including machine learning and deep learning models, are used with clinical, radiological, and histopathological data. These AI methods were compared against conventional staging systems and clinical approaches for head and neck cancer. The main goal was to see if AI could improve risk stratification, specifically predicting lymph node metastasis and extranodal extension. The review found that the predictive performance of these AI models ranged from moderate to high. No safety concerns were reported because these are computer-based tools rather than treatments given to patients. However, the studies had significant limitations. There was substantial methodological heterogeneity, meaning the studies were not done in the same way. Most designs were retrospective, and there was limited external validation of the models. The review also noted insufficient assessment of the actual clinical impact of these proposed models. While AI has the potential to enhance risk stratification and complement conventional approaches, it is not yet ready for routine use. Readers should understand that methodological standardization, prospective multicentre validation, model interpretability, and ethical issues still need work before these tools can be widely adopted.