ESTRO 2025 - Abstract Book

S2538

Physics - Autosegmentation

ESTRO 2025

Figure 2: Box plots of DSC coefficients for ATLAAS and DeepLabV3+-based method on the PEARL test set.

Conclusion: We propose a novel two-stage method for GTV delineation in PET/CT images of HNC, achieving comparable median DSC performance to the ATLAAS method. Our approach shows promise for consistent and automated segmentation in radiotherapy planning.

Keywords: HNC, Deep Learning, PET/CT

References: [1] Andrearczyk, V., et al (2021). https://doi.org/10.1007/978-3-030-98253-9_1. [2] Chen, L.C., et al (2018). https://doi.org/10.1007/978-3-030-01234-2_49. [3] Myronenko, A., et al (2022). https://doi.org/10.1007/978-3-031-27420-6_2. [4] Berthon, B., et al (2016). https://doi.org/10.1088/0031-9155/61/13/4855. [5] Svobodova, M., et al (2018). https://www.ncri.org.uk/abstract/pearl-pet-based-adaptive-radiotherapy-clinical trial

3720

Proffered Paper Accuracy, multi-center transferability, and usability of interactive deep-learning for head and neck gross tumor volume segmentation Zixiang Wei 1,2 , Jintao Ren 1,2 , Stine S Korreman 1,2,3 , Jesper Grau Eriksen 1,3,4 , Kenneth Jensen 2 , Hanna Rahbeck Mortensen 2,1 , Zeno Gouw 5 , Dede Cem 6 , Kareem A Wahid 6 , Jan-Jakob Sonke 5 , Clifton Dave Fuller 6 , Jasper Nijkamp 1,2 1 Clinical medicine, Aarhus University, Aarhus, Denmark. 2 Danish Center for Particle Therapy, Aarhus University Hospital, Aarhus, Denmark. 3 Oncology, Aarhus University Hospital, Aarhus, Denmark. 4 Experimental Clinical Oncology, Aarhus University Hospital, Aarhus, Denmark. 5 Radiation Oncology, Netherlands Cancer Institute, Amsterdam, Netherlands. 6 Radiation Oncology, University of Texas MD Anderson Cancer Center, Houston, USA