Radiomics and Radiogenomics in Prognostic Assessment of Head and Neck Cancer: A Systematic Review of Cutting-Edge...

The integration of radiomics and radiogenomics in the prognostication of head and neck cancer represents a rapidly evolving field within precision oncology. This systematic review aims to appraise advanced methods in radiomics and radiogenomics concerning prognostication in head and neck cancer, with a particular focus on methodological developments and clinical applications.

A systematic literature search was conducted across seven major electronic databases: PubMed/MEDLINE, Embase, Web of Science, Scopus, Cochrane Library, CINAHL, and Google Scholar from January 2013 to December 2023. The search strategy incorporated database-specific syntax, controlled vocabulary such as Medical Subject Headings (MeSH) and Emtree terms, and supplementary free-text terms.

Twelve studies were included in the review, and the qualitative analysis revealed three distinct research clusters: prognostic applications, development of predictive models, and molecular-immunological characterization. In all scenarios, studies employing multi-modality modelling were significantly more competent than those relying on single-modality analyses. The area under the curve values of machine learning ranged across 0.71-0.86, outperforming traditional statistical approaches. Larger cohort studies exhibited superior validation metrics. While predictions of molecular characteristics varied, the prediction of immune phenotypes was superior to that of specific genetic alterations. Studies incorporating external validation provided stronger evidence supporting clinical usability. Although some studies presented moderate risk due to early-phase methodological variability, nearly half demonstrated low overall bias.

Our findings indicate significant advances in the prognostication of head and neck cancer through radiomics and radiogenomics approaches. Combined modelling strategies that integrate clinical, radiomic, and genomic features yielded enhanced performance. Despite the development of newer studies with greater methodological rigor and robust validation, variability in feature extraction, processing pipelines, and reporting metrics necessitates further consolidation of methodologies in this field.