S2121
Physics - Inter-fraction motion management and daily adaptive radiotherapy
ESTRO 2026
Figure 1 illustrates inter-observer variability relative to the reference contours, while figure 2 summarizes DSC and MDA across structures. Overall, DSC and MDA analyses demonstrated acceptable agreement. The target and great vessels showed significantly larger inter-observer and inter-fraction variability (p < 0.05), underscoring persistent challenges in delineating irregular or elongated structures. Notably, online GTV contours tended to be larger than Ref contours, implying a more conservative approach whenever users considered several anatomical regions doubtful. The bile duct also varied between observers, although this did not reach statistical significance, likely due to its small size. In contrast, larger and well-defined organs such as the stomach and duodenum exhibited consistently high concordance and low spatial deviations. GI substructures (e.g., colon, jejunum) showed wide inter- observer variation individually; however, the composite GI volume remained consistent (p > 0.05), supporting the clinical practice of applying common dose constraints to combined GI organs. Conclusion: Daily adaptive MR-guided SBRT for pancreatic cancer patients, performed online by different observers, demonstrates acceptable delineation consistency. Although limited, variability was significantly more pronounced for small and complex structures, such as the GTV, great vessels, and bile duct, while the composite total GI volume remained stable across observers. Nevertheless, further dosimetric evaluation is needed to assess the clinical impact of these OAR variations. Keywords: Pancreatic cancer, Observer Variability, MRgSBRT Proffered Paper 1131 Imaging for adaptive particle therapy - A survey of the European Particle Therapy Network Lena Nenoff 1,2 , Pietro Pisciotta 3 , Jenneke Jacobs 4 , Kristyna Olsson Hermanova 5,6 , Petra Trnkova 5,7 , Alexandru Dasu 8,9 , Ulrik Vindelev Elstrøm 10 , Grete May Engeseth 11 , Callum Gillies 12 , Mischa Hoogeman 4 , Dominika Kedzierska-Pardel 13 , Friderike K Longarino 14,15 , Hanna Rahbek Mortensen 10,16 , Sandija Plaude 17 , Mirko Unipan 18 , Lamberto Widesott 19 , Joe Wood 20 , Andrea Zechner 21 , Tony Lomax 22 , Markus Stock 21,23 , Alessandra Bolsi 22 , Christian Richter 1,24 , Francesca Albertini 22 1 OncoRay – National Center for Radiation Research in Oncology, Faculty of Medicine, Dresden, Germany. 2 Helmholtz-Zentrum Dresden-Rossendorf, Institute of Radiooncology – OncoRay, Dresden, Germany. 3 University Medical Center Groningen UMCG, Department of Radiation Oncology, Groningen,
Purpose/Objective: Pancreatic cancer is well suited for MR-guided SBRT (MRgSBRT) due to superior soft-tissue contrast and daily adaptation, especially with isotoxic OAR-sparing dose prescription. Still, online adaptations are time- limited and often performed by different physicians, introducing inter-observer variability and potential uncertainty in treatment consistency.The aim of this study is to evaluate the potential impact of inter- observer variability for pancreatic cancer patients undergoing MRgSBRT. Material/Methods: Ten pancreatic cancer patients treated on the 1.5T MR- Linac between July 2023 and October 2024 with a dose prescription of 50Gy in 5 fractions were retrospectively analyzed. For each of the 50 fractions, target and adjacent OARs were delineated by different physicians [“online”] followed by offline independent contouring by a single expert physician [“ref”]. OARs were evaluated within a 3 cm region-of-interest (ROI) surrounding the target. Agreement between online and reference contours was assessed using the Dice Similarity Coefficient (DSC) and Mean Distance to Agreement (MDA) for the target [gross tumor volume (GTV) and tumor–vessel interface (TVI)], gastrointestinal organs (GI), bile duct, and great vessels. Statistical analysis was performed using “Student t-test”, at a significance level of p < 0.05 using SciPy python package (version 1.5.2). Results:
Made with FlippingBook - Share PDF online