Abstract / Summary
Importance: Artificial intelligence (AI) is increasingly integrated into clinical decision-making, including applications in pathology, radiomics, and multidisciplinary tumor boards (MDT) for head and neck (ENT) cancer. However, the current scope and performance of AI tools within MDT settings remain unclear.
Objective: To systematically evaluate the existing literature on the use and performance of AI models in MDT decision-making for ENT cancer.
Design/setting: Systematic review.
Participants: Studies assessing AI applications in MDT for head and neck (ENT) cancer.
Interventions: A systematic search of Cochrane, Embase, Medline, PubMed, and Scopus was conducted from database inception to March 4, 2025. Studies were included without language or date restrictions. The review was conducted in accordance with PRISMA guidelines.
Main outcomes measures: AI performance across 4 key domains: therapy option selection, clinical recommendations, explanation capabilities, and summarization abilities.
Results: Of 791 records screened, 37 underwent full-text review and 7 met inclusion criteria. Six AI models were evaluated, with ChatGPT-4 featured in 71% of included studies. Across assessed domains, ChatGPT-4 demonstrated superior performance compared to other models, particularly in therapy selection, recommendation quality, and explanatory capacity.
Conclusions: The current evidence base regarding AI integration in MDT decision-making for ENT cancer remains limited and heterogeneous. Although ChatGPT-4 shows promising performance across multiple domains, concerns regarding clinical accuracy, consistency, and transparency persist. Further research is required to optimize AI reliability and validate its clinical utility.
Relevance: Clarifying the capabilities and limitations of AI within MDT settings is essential to ensure safe integration into oncologic decision-making and to support future development of reliable, transparent AI-assisted clinical tools.