S2132
Physics - Inter-fraction motion management and daily adaptive radiotherapy
ESTRO 2026
Erasmus MC Cancer Institute, Department of Radiotherapy, Rotterdam, Netherlands
Purpose/Objective: This retrospective study investigates whether advanced image registration methods for dose accumulation are necessary to fully exploit the potential of fractionated online adaptive SBRT for lymph node oligometastases. Specifically, it examines if these methods improve prediction for in-treatment dose adaptation, where fraction dose is increased isotoxically in case of favorable anatomy to reduce the number of treatment fractions. Material/Methods: Twenty-five patients with abdominal-pelvic lymph node oligometastases from a phase II study (STEAL, NL58442.078.17) were included. Each patient had a planning CT and five fraction CTs (fCT) with delineated segments of gastrointestinal organs (GIOs). Online adaptive treatments of 5x9 Gy were simulated, GIO delineations were propagated, and doses were accumulated using five different techniques: DVH summation, rigid registration (MIM-Rigid), image- based & hybrid deformable registrations (MIM-DIR- Image & MIM-DIR-Hybrid), and Thin Plate Spline Robust Point Matching (TPS-RPM). Each fCT was alternately treated as a reference to assess robustness (Figure 1). Geometric accuracy was evaluated using Dice coefficient (DSC), Hausdorff 99th percentile (HD- 99%), and mean surface distance (MSD). Accumulated dose metrics included V20Gy3 (EQD2) and D 0.5cc. The best-performing robust algorithm was used in the in- treatment dose adaptation workflow and the reduction in fractions compared to DVH summation. In this workflow, each fraction was replanned isotoxically aiming to deliver the remaining treatment dose and reduce fractions.
Figure 1. Qualitative assessment results.
Figure 2. Examples illustrating improved contour propagation toward the lungs, heart, and aorta when their binary masks were provided to the DIR algorithm.Treatments based on unadjusted contours from contour-guided DIR achieved high target coverage, with a minimum planning CTV V95% of 97.8% and the remaining four patients exceeding 99%. Conclusion: Incorporating contour information into DIR improved esophageal target propagation from the planning to fraction MR scan. Consequently, our contour-guided DIR can improve the accuracy and efficiency of online adaptive MR-guided radiotherapy in esophageal cancer. These findings also suggest that these target contours could potentially be used without manual adjustments or additional margin expansions. References: 1. Isensee, F., Jaeger, P. F., Kohl, S. A. A., Petersen, J., & Maier-Hein, K. H. (2021). nnU-Net: a self-configuring method for deep learning-based biomedical image segmentation. Nature Methods, 18(2), 203–211. https://doi.org/10.1038/s41592-020-01008-z Keywords: Contour propagation; Esophageal CTV; MR- guided RT
Results: Geometric accuracy improved significantly as registration complexity increased (Table 1). TPS-RPM achieved the highest geometric accuracy (median DSC 0.98, HD-99% 1.4 mm, MSD 0.25 mm) and lowest variability. TPS-RPM was the only method to predict a significantly lower accumulated GIO D0.5cc compared to DVH summation (25.5 Gy vs 26.9 Gy). V20Gy3 for TPS-RPM and MIM-DIR-Hybrid was also significantly lower than DVH summation (5.4 Gy and 5.2 Gy vs 5.9
Mini-Oral 2067
Do we need advanced image registration for dose accumulation to make decisions in online adaptive SBRT? Erik van Lieshout, Mischa S. Hoogeman, Remi A. Nout, Joost J.M.E. Nuyttens, Maaike T.W. Milder
Made with FlippingBook - Share PDF online