Application of radiomics in head and neck squamous cell carcinoma

• Radiomics models show promise in predicting clinically relevant factors in HNSCC, including HPV status, lymph node metastasis/extranodal extension, and survival endpoints (OS/PFS/RFS). • Delta-radiomics and machine learning approaches can support early prediction of radiotherapy-related toxicities (e.g., xerostomia, trismus, dysphagia), facilitating personalized treatment planning. • Radiogenomics and radioproteomics link imaging phenotypes with molecular profiles (e.g., somatic mutations and immune biomarkers such as PD-L1), providing a foundation for imaging-based multi-omics in HNSCC. • Key barriers to clinical translation include feature standardization, acquisition/segmentation variability, and limited external validation; multicenter prospective studies and harmonization strategies are priorities for future research. Head and neck squamous cell carcinoma (HNSCC) is the most prevalent histopathological subtype of head and neck malignancies. Owing to the lack of early specific symptoms, the majority of patients are diagnosed at intermediate or advanced stages, which is associated with an unfavorable prognosis and a substantial decline in quality of life. Radiomics, which leverages large-scale medical imaging data, enables the extraction of high-dimensional quantitative features through advanced image analysis, thereby providing deeper insights into tumor biology. In this review, we summarize recent advances in radiomics for the diagnosis, prognostic prediction, and evaluation of treatment-related toxicity in HNSCC. Furthermore, we highlight emerging applications of radiomics in genomics and proteomics, illustrating the associations between tumor molecular phenotypes and imaging-derived features. Finally, we discuss the challenges related to feature standardization and reproducibility, and outline the key limitations of current radiomics studies.