Dose-based evaluation of delineation variation in radiotherapy: A scoping review

Delineation of targets and organs at risk (OARs) is necessary for radiation therapy (RT) treatment, yet remains prone to inter-observer variability (IOV). Auto-segmentation is often used, however an underlying challenge for both auto-segmentation and IOV studies is assessing delineation quality. While geometric metrics are widely used, they provide limited clinical insight. Ideally, evaluation should reflect clinical impact, though direct outcome-based assessment is rarely feasible. Dose-based evaluation, which assesses how delineation differences affect delivered or reported dose, offers a potential surrogate to clinical relevance. This scoping review provides a comprehensive overview of dose-based evaluation for delineation studies in RT, including auto-segmentation and IOV, published before December 2024. Following PRISMA-ScR guidelines, two independent reviewers searched two databases (PubMed, Scopus) and screened manuscripts, yielding 144 eligible studies. We observed substantial heterogeneity in evaluation approaches, including methods, metrics and definition of clinically meaningful dose differences. Most studies reported no or minor dosimetric impact for OARs (84%), though within studies, notable outlier cases often existed. Negligible dose differences were less frequently reported for targets (31%). Factors influencing sensitivity to geometric delineation variations, such as proximity to high-dose gradients or structure size, were found to be context-dependent. Clinical and planning context, including dose conformity, was reported to strongly influence the effect of delineation variations on dose. Overall, the review highlights high methodological heterogeneity and underscores the need for careful consideration of the steps involved in dose-based evaluation. Improving comparability and interpretability of dose-based evaluation requires standardisation of frameworks and consensus on the definition of clinically meaningful dose differences.