S1705
Physics - Dose prediction/calculation, optimisation and applications for particle therapy planning
ESTRO 2026
Two clinical treatment fields with gantry-equivalent angles of 210° (G210) and 270° (G270) were also delivered. The G270 field has previously served as reference case in [3] for benchmarking the PGI (iba Knife-Edge Slit Camera) and the PGS (Massachusetts General Hospital) systems.Geometric range shifts of 2.2–8.9 mm were introduced by inserting PMMA plates with thicknesses of 2, 4, 6, and 8 mm in front of the phantom. To increase statistical robustness, all measurements were repeated multiple times. For each spot, PGT distributions were extracted and processed, and 62 features were computed and normalized relative to corresponding reference measurements without additional plates.A multi-step machine learning workflow was implemented for feature extraction and model development. Relevant predictors for range shift estimation were selected, and a linear regression model was trained on the synthetic field data. The final feature set was defined based on model performance on the G210 validation field, and unbiased model testing was performed on the G270 field. Evaluation metrics were applied as described in [3]. Results: On the final test data (G270), the model achieved an absolute range accuracy of − 0.32 ± 0.03 mm, an absolute range precision of 0.88 mm, and a mean spot-wise precision of 1.63 mm.For the same phantom and irradiation plan, these results are consistent with those obtained using the iba Knife-Edge Slit Camera (PGI) and demonstrate improved accuracy compared to the PGS system from MGH, albeit with slightly reduced precision (Table 1). Conclusion: The developed PGT-based model demonstrated range-shift prediction performance comparable to that of PGI and PGS systems in clinically realistic conditions, confirming the potential of PGT as a viable tool for proton range verification. References: [1] Kieslich A, Schellhammer SM, Zwanenburg A, Kögler T, Löck S, et al. Machine-learning-based integration of temporal and spectral prompt gamma- ray information for proton range verification. Phys Imaging Radiat Oncol. 2025;35:100788.[2] Schellhammer SM, Wiedkamp J, Löck S, Kögler T. Multivariate statistical modelling to improve particle treatment verification: Implications for prompt gamma-ray timing. Front Phys. 2022;10:932950.[3] Hueso-González F, Berthold J, Wohlfahrt P, Bortfeld T,
Poster Discussion 773
Benchmarking prompt gamma-ray timing for proton range verification in clinically realistic experiments Aaron Kieslich 1,2 , Kristina Makarevich 1,2 , Alex Zwanenburg 1,3 , Toni Kögler 1,2 , Steffen Löck 1,2 1 OncoRay—National Center for Radiation Research in Oncology, Faculty of Medicine and University Hospital Carl Gustav Carus, TUD Dresden University of Technology , Helmholtz-Zentrum Dresden – Rossendorf, Dresden, Germany. 2 Helmholtz-Zentrum Dresden - Rossendorf, Institute of Radiooncology – OncoRay, Dresden, Germany. 3 National Center for Tumor Diseases Dresden (NCT/UCC), Germany: German Cancer Research Center (DKFZ), Heidelberg, Germany; Faculty of Medicine and University Hospital Carl Gustav Carus, TUD Dresden University of Technology, Dresden, Germany; Helmholtz-Zentrum Dresden-Rossendorf (HZDR), Dresden, Germany Purpose/Objective: Prompt gamma-ray timing (PGT) is a promising approach for in-vivo range verification in proton therapy [1,2]. This study aims to develop a PGT-based range-shift prediction model under clinically realistic conditions and to evaluate its performance within a benchmark framework for prompt gamma-ray systems [3], enabling a direct comparison of PGT to prompt gamma-ray imaging (PGI), and prompt gamma-ray spectroscopy (PGS) under similar experimental conditions. Material/Methods: An anthropomorphic head phantom (CIRS 731-HN) was irradiated with two clinical and multiple synthetic irradiation fields (Figure 1). Synthetic fields were delivered at nominal energies of 139, 162, and 187 MeV in a 15 × 15 spot grid (225 spots) with 0.2 MU or 1 MU per spot, applied at a gantry-equivalent angle of 180°, corresponding to a dorsal beam incidence.
Made with FlippingBook - Share PDF online