S1845
Physics - Dose prediction/calculation, optimisation and applications for photon and electron planning
ESTRO 2026
2025 Keywords: TCP, Probabilistic Planning, CTV
TCP. Nevertheless, enforcing the invariance of expected tumor load and TCP to voxel size, we derived the following: This new TCP formulation is invariant to voxel geometry due to the probability expressed as a cell density and will be the closest to Monte Carlo simulations (Figure 2). Because of the convexity of its log-form and its conservative approximation of reality, this model is the ideal candidate for both TCP-based treatment plan optimization and evaluation2.
Digital Poster 1579 Automated VMAT Planning for Ultra- Hypofractionated Prostate Radiotherapy Pauline Maury 1 , Willliam Green 2 , Rafe McBeth 2 , Enis Ozyar 3 , Gorkem Gungor 3 , Madalina-Liana Costea 4 , Baris Ungun 5 , Remi Vauclin 5 , Elie Mengin 5 , Nikos Paragios 6 , Mario Terlizzi 1 1 Radiation Oncology Department, Institute Gustave Roussy, Villejuif, France. 2 Radiation Oncology Department, University of Pennsylvania, Philadelphia, USA. 3 Radiation Oncology Department, Acibadem MAA University School of Medicine, Istanbul, Turkey. 4 Clinical Affairs, TheraPanacea, Paris, France. 5 Physics, TheraPanacea, Paris, France. 6 CEO, TheraPanacea, Paris, France Purpose/Objective: Ultra-hypofractionated radiotherapy is an established treatment for localized prostate cancer [1]. This study aimed to clinically validate an automated planning (AP) pipeline combining deep learning-based dose prediction with dose mimicking to efficiently generate high-quality, clinically acceptable VMAT plans. Material/Methods: The AP workflow was trained on 138 manually generated VMAT plans (dose prescription: 36.25 Gy in 5 fractions). Clinical validation was conducted on 20 patients from two European centers. Using planning CTs and validated target and OAR contours, the AP pipeline automatically produced VMAT treatment plans (11 for VersaHD with Agility MLC, 9 for TrueBeam with HDMLC), and corresponding doses were calculated in RayStation and Eclipse Treatment Planning Systems. A quantitative comparison between AP and manual plans (MP) was done following GETUG recommendations [1], with all AP plans normalized to 100% CTV coverage. Statistical analysis was performed using the Wilcoxon rank test. Three senior radiation oncologists (two European, one US) performed independent qualitative evaluation, with freedom to adjust normalization and assess overall clinical acceptability. A subset of cases underwent patient- specific QA to confirm deliverability. Results: The full AP process required approximately 10 minutes, including dose calculations. With an additional ~10 minutes for physician review, clinically ready plans were available within ~20 minutes. AP plans demonstrated superior target coverage compared to MPs (Table 1), achieving CTV coverage in 20/20 cases versus 9/20 for MPs (p<0.001). MPs showed slightly improved rectum D1cc performance
Conclusion: Probabilistic target volumes rely on the fundamental hypothesis of voxel correlation, which, due to the voxel geometry, will impose a bias on tumor expansion and thereby on TCP. By carefully reformulating the definition of the probability mapping step, we derived a state-of-the-art analytical formulation for an unbiased expected TCP. References: [1] G. Butiz et al., «Exploring trade - offs in treatment planning for brain tumor cases with a probabilistic definition of the clinical target volume» 2023[2] E. Peeters, R. Schyns «Exploring the Computational and Clinical Added-value of Probabilistic Optimization in Proton Therapy» 2024[3] T. Bortfeldz et al.,«Probabilistic definition of the clinical target volume—implications for tumor control probability modeling and optimization» 2021[4] X. Meng et al.,«Noninvasive Evaluation of Microscopic Tumor Extensions Using Standardized Uptake Value and Metabolic Tumor Volume in Non-Small-Cell Lung Cancer» 2012[5] L. Amoudruz et al., «Ising energy model for the stochastic prediction of tumor islets»
Made with FlippingBook - Share PDF online