S2780
RTT - RTT contouring, target definition, and treatment planning
ESTRO 2026
Martin Otto, Maarten Dirkx, Erik Lieshout, Danielle van Nispen, Maaike Milder, Joost Nuyttens Radiotherapy, ErasmusMC Cancer Institute, Rotterdam, Netherlands
an NVIDIA A40 GPU. All optimized plans met clinical treatment criteria for target coverage and organ at risk constraints.
Purpose/Objective: Until 2024 patients with abdominal-pelvic
oligometastases were treated at the department of Radiotherapy at the Erasmus MC Cancer Institute, with five fractions of 9 Gy using the CyberKnife system (Accuray Inc.). However, due to inter-fraction organ motion in some fractions either the dose coverage of the target volume was insufficient or the maximum dose constraint in an organ at risk (OAR) (D0.5cc<35Gy) was violated. The CyberKnife system does not offer a solution to correct for these violations for instance by adapting the dose to the daily anatomy. Hence, we investigated the possibility of using the online adaptive workflow offered by the Ethos system (Varian Medical System Inc.) as an alternative in the STEAL-3 (NL87228.078.24) trial. This trial required the development of SBRT planning templates for the Ethos system, a first in our institute. Due to vicinity of the OAR to the target volume, different planning strategies had to be developed to obtain plans with sufficient quality. Material/Methods: In the Ethos treatment planning system (TPS) planning templates were designed. Pre-fraction, repeat CT scans of patients previously treated at the CyberKnife were used for simulation and evaluation of the online adaptive workflow. A 3mm PTV margin was used. The high-fidelity mode in the Ethos TPS was enabled. Target coverage was compromised to obey OAR constraints. Results: To respect the dose constraints of the OAR, a 1 or 2 mm PRV margin to control the D0.5cc<35Gy and Dmax was required. A ring structure of PTV minus GTV was created to prevent high dose spill outside the target volume. However, the initial Ethos plans lacked the same dose heterogeneity compared to the CyberKnife plans. To achieve higher dose heterogeneity a goal of D0.03cc>125% was added to the target volume. For larger volume targets adding a D118%> goal improved dose heterogeneity, to resemble CyberKnife plans. In the ongoing STEAL-3 trial dose constraints of the OAR were not violated, due to online dose re-optimization. Conclusion: The online adaptive workflow at Ethos therapy offers the possibility of treating patients with abdominal- pelvic oligometastases with adaptive SBRT treatment plans, solving dose constraint violations. Ethos TPS achieved similar plan quality compared to CyberKnife plans with the created SBRT planning templates. Based on the first patients treated in the STEAL-3 trial,
Conclusion: This physics-informed framework provides the foundation for end-to-end optimization from patient images to machine parameters, decoupled from TPS infrastructure. Embedding dose calculation within the computational graph allows gradients to flow through physical dose transport, connecting hardware-aware, deterministic modeling and DL optimization. A companion abstract (Glatzer et al.) demonstrates its application in direct VMAT plan prediction[2], highlighting its translational potential for adaptive and preference-driven planning. The PyDoseRT package will be released as an open-source research tool, enabling reproducible, physics-grounded AI development in radiotherapy. References: [1] Tufve Nyholm, Jörgen Olofsson, Anders Ahnesjö und Mikael Karlsson. „Photon pencil kernel parameterisation based on beam quality index“. In: Radiotherapy and Oncology 78 (3 March 2006), S. 347– 351. issn: 01678140. doi: 10.1016/j.radonc.2006.02.002[2] Gerd Heilemann, Lukas Zimmermann, Raphael Schotola, Wolfgang Lechner, Marco Peer, Joachim Widder, Gregor Goldner, Dietmar Georg und Peter Kuess. „Generating deliverable DICOM RT treatment plans for prostate VMAT by predicting MLC motion sequences with an encoder-decoder network“. In: Medical Physics 50 (8 August 2023), S. 5088–5094. issn: 24734209. doi: 10.1002/mp.16545 Keywords: deep learning, treatment planning, dose engine
Digital Poster 2340
Online adaptive SBRT treatment of abdominal- pelvic oligometastases: from Cyberknife to Ethos
Made with FlippingBook - Share PDF online