Predicting post-radiotherapy dysphagia in head and neck cancer: A narrative review of emerging artificial intelligence models

Abstract

Head and neck cancer (HNC) remains a significant global health burden, particularly due to the high incidence of treatment-related complications such as dysphagia. Post-radiotherapy dysphagia not only impairs swallowing and nutritional status but also reduces quality of life and increases healthcare utilization. Conventional diagnostic approaches to identify patients at risk are often limited by delayed detection, variability in assessment methods, and insufficient predictive accuracy. Predictive modeling, including statistical, machine learning, and deep learning approaches, has emerged as a promising strategy to anticipate dysphagia risk before or during treatment. Techniques such as radiomics, imaging-based biomarkers, and multimodal machine learning models allow for early identification of high-risk patients, thereby enabling tailored interventions and preventive strategies. The aim of this review is to map the current landscape of predictive models for post-radiotherapy dysphagia in HNC patients, highlighting methodologies, applications, and clinical implications. A comprehensive literature search was performed across PubMed, Scopus, Web of Science, and Embase databases, including studies published up to February 3, 2025. Evidence indicates that predictive models, particularly those integrating machine learning and deep learning approaches, show potential in improving accuracy and timeliness of dysphagia risk prediction. However, challenges remain, including heterogeneity of patient data, limited external validation, and barriers to clinical integration. This review highlights the potential of predictive modeling to enhance individualized care in HNC patients and emphasizes the need for standardized methodologies, multicenter validation, and real-world implementation to optimize outcomes.