ESTRO 2025 - Abstract Book

S2492

Physics - Autosegmentation

ESTRO 2025

Conclusion: We identified the introduction of automation bias, an overall decrease in editing with improved auto-segmentation solutions and an unexpected trend for the oral cavity, which warrants further investigation on its clinical impact and possibly an update of our deep learning model for this structure. The findings showcase the relevance of having a monitoring system in place following the implementation of auto-segmentation.

Keywords: manual editing, continuous monitoring

References: [1] Vandewinckele L, Claessens M, Dinkla A, Brouwer C, Crijns W, Verellen D, et al. Overview of artificial intelligence based applications in radiotherapy: recommendations for implementation and quality assurance. Radiother Oncol 2020;153:55-56 [2] Hurkmans C, Bibault J-E, Brock KK, van Elmpt W, Feng M, Fuller CD, et al. A joint ESTRO and AAPM guideline for development, clinical validation and reporting of artificial intelligence models in radiation therapy, Radiother Oncol 2024;197 [3] Vaassen F, Hazelaar C, Vaniqui A, Gooding M, van der Heyden B, Canters R, et al. Evaluation of measures for assessing time-saving of automatic organ-at-risk segmentation in radiotherapy. Phys Imaging Radiat Oncol 2020;13:1 6

2761

Digital Poster TotalSegmentator in radiation oncology: a validation study on clinical imaging data Maksym Fritsak 1,2 , Hubert S. Gabryś 1 , Matthias Guckenberger 1,2 , Stephanie Tanadini-Lang 1 1 Department of Radiation Oncology, University Hospital Zurich and University of Zurich, Zurich, Switzerland. 2 Faculty of Medicine, University of Zurich, Zurich, Switzerland Purpose/Objective: In radiation oncology research, big data analysis plays a crucial role in advancing the understanding of treatment outcomes and optimizing therapeutic strategies. A fundamental aspect of such analyses is the segmentation of normal anatomical structures, however, it is time-consuming, labor-intensive, and impractical for large datasets.