ESTRO 2025 - Abstract Book

S2780

Physics - Dose prediction, optimisation and applications of photon and electron planning

ESTRO 2025

dose distribution to the primary. This study provides a useful methodology and recommendations for medical physicists to evaluate and limit fetal dose, which should be kept as low as reasonably achievable, while maintaining treatment quality to the primary.

Keywords: pregnant woman, fetal dose reduction

References: [1] C. Maggen et al. , « Pregnancy and Cancer: the INCIP Project. », Curr. Oncol. Rep. , vol. 22, n o 2, p. 17, févr. 2020, doi: 10.1007/s11912-020-0862-7 [2] ICRP Publication 84, « Pregnancy and Medical Radiation. » Ann. ICRP 30, 2000. [3] M. Michalet, C. Dejean, U. Schick, C. Durdux, A. Fourquet, et Y. Kirova, « Radiotherapy and pregnancy. », Cancer Radiother. J. Soc. Francaise Radiother. Oncol. , vol. 26, n o 1-2, p. 417-423, avr. 2022, doi: 10.1016/j.canrad.2021.09.001. [4] A. Chaikh et al. , « Modeling and dosimetric characterization of a 3D printed pregnant woman phantom for fetal dosimetry in radiotherapy », Radioprotection , oct. 2024, doi: 10.1051/radiopro/2024039.

1798

Mini-Oral Automating High-Dimensional IMRT Plan Optimization Using an Advanced Bayesian Method Vinay Saini 1,2 , Neeraj Sharma 2 , Satyajit Pradhan 1 , Abhishek Shinghal 1 , Sanjay Barman 1 , Alok Kumar Srivastava 2 1 Department of Radiation Oncology, Mahamana Pandit Madan Mohan Malaviya Cancer Centre & Homi Bhabha Cancer Hospital, Tata Memorial Centre, Homi Bhabha National Institute, Varanasi, India. 2 School of Biomedical Engineering, Indian Institute of Technology (BHU), Varanasi, India Purpose/Objective: Traditional inverse planning for radiation therapy (RT) predominantly includes manual adjustments of planning parameters like dose objectives and weightage values. This complex and time-consuming manual process results in suboptimal plans because of variable experience of planners and time-constraints. Standard Bayesian Optimization (SBO) methods, although effective in management of up to 20 parameters, sometimes have issues in high dimensional spaces that are common in RT planning, where the number of parameters often exceeds 20. The present study investigates the application of a recently updated advanced SBO method, with capability for dynamically adjusting model complexity with increasing dimensions. This technique supports consistent performance across complex optimization landscapes and is particularly effective in high-dimensional spaces encountered in RT. The refined approach aims to streamline the search for global solutions and enhance the automated optimization of intensity-modulated radiation therapy (IMRT) planning. Material/Methods: The present retrospective study adapted an IMRT planning auto-optimization script, developed in Python, to utilize the advanced SBO method, and integrated it into Varian’s Eclipse TPS v18.0 and focused it on cancer treatment plans. The optimization process targeted dose objectives and their respective weightages, ensuring compliance with the EMBRACE-II protocol. The study was performed on plans of 15 cancer cervix patients treated earlier with simultaneous-integrated boost (SIB) that delivered 55 Gy/25 fractions to lymph nodes and 45 Gy/25 fractions to primary tumor and elective targets. Dose-volume histogram (DVH) metrics were used for comparative analysis between SBO-generated plans and manually created expert clinical plans. Results: SBO-generated plans achieved comparable target coverage and organ sparing to those created by expert planners. While key parameters such as GTVn(D98%) and ITV45(D99.9%) showed no significant differences (p>0.05), notable improvements were observed in dosimetric indices for the femur, kidney, rectum, bladder, and bowel (Figure-1 and Figure-2). The enhancement in organ sparing and overall plan conformity highlight the future potential of advanced