S2177
Physics - Inter-fraction motion management and daily adaptive radiotherapy
ESTRO 2026
Conclusion: Differences between configurations remained moderate, with greater variation observed between the two phantoms than between software programs, suggesting that DIR performance is mainly constrained by image quality and anatomical variability rather than by algorithmic settings. Adjusting the VOI to a smaller regions improved metrics for all software programs. Among the tested configurations, Velocity with reduced-VOI-Multipass and AutoContour-reduced-VOI yielded the most robust and consistent outcomes. These findings provide a practical framework for benchmarking DIR performances and could help other institutions evaluate their own DIR algorithms. Further validation on patient datasets is required to confirm the optimal settings for clinical implementation in adaptive and re-irradiation workflows. Keywords: Deformable image registration Development and implementation of online adaptive radiotherapy (oART) using a Radixact delivery platform in combination with the RayStation TPS Alex Dunlop 1 , Michael Thomas 1 , Shabanaz Boodhoo 2 , Irena Blasiak-Wal 1 , Emily Hogg 1 , Susan Lalondrelle 3 , Benjamin Thomas 3 , Ryan Fullarton 4 , Simeon Nill 1 , Uwe Oelfke 1 1 The Joint Department of Physics, The Institute of Cancer Research and The Royal Marsden Hospital, London, United Kingdom. 2 Department of Radiotherapy, The Royal Marsden Hospital, London, United Kingdom. 3 Department of Radiotherapy and Imaging, The Institute of Cancer Research and The Royal Marsden Hospital, London, United Kingdom. 4 Radiotherapy Physics, University College London Hospitals NHS Foundation Trust, London, United Kingdom Digital Poster 4265 Purpose/Objective: With adequate imaging [1] combined with the ability to image and treat long fields, the Radixact (Accuray, Wisconsin) treatment platform is ideal for X-ray-based online adaptive radiotherapy (oART). In this work we developed and clinically implemented an oART workflow for Radixact using the RayStation (RaySearch Laboratories, Stockholm) TPS (v2024B onwards). Material/Methods: Existing tools within the TPS were evaluated and commissioned for the online workflow including: corrected CBCT conversion for more accurate dose calculation; rigid and deformable contour propagation from a reference image and AI-based auto-contouring directly on the daily corrected image set; automated replanning functionality available within the software.
Results: Across all software programs, contour propagation accuracy varied by organ and DIR parameters (Figure 2). The TG132-DP rectum consistently demonstrated the highest accuracy, with DSC values up to 0.96, likely due to its higher contrast and stable volume. In contrast, the RUBY-Adpt bladder showed the lowest DSC values (down to 0.57), due to poor contrast with surrounding tissues and greater volume variability.For the RUBY-Adpt, overall DIR performance was lower across all metrics compared with the TG132-DP, likely due to reduced image contrast. Nevertheless, reducing the DIR VOI generally improved registration accuracy.In Velocity, comparison between the Multipass and Extended modes showed no consistent advantage for the Extended mode. The reduced-VOI- Multipass yielded the most stable and accurate results (e.g., DSC 0.71–0.90, MDA 1.27–4.19 mm, HDmax 6.0– 17.0 mm).
Made with FlippingBook - Share PDF online