ESTRO 2025 - Abstract Book

S4328

RTT - Treatment planning, OAR and target definitions

ESTRO 2025

Material/Methods: The planning Computed Tomography (CT) and Organs at risk (OARs) structures of 40 patients with left breast cancer were used as training data to build an auto segmentation model. Based on the developed model using Deep Learning (DL) algorithm, auto contours were generated on a separate group of planning CTs of 30 left breast cancer patients. The accuracy and feasibility were assessed through four evaluations including geometric evaluation, dose evaluation, time evaluation and user satisfaction evaluation by RTTs. Results: The auto segmentation model achieves over 0.7 in DSC except left anterior descending artery (LAD). 95% HD of all OARs are within 40mm. The dose for the treatment plans generated from manual contours and auto contours were found to have no significant differences. With the assistance of auto contours, the contouring time was reduced by 60%. The performance scores of all auto contours except LAD and stomach were above 3 meaning only minor adjustments were required. Conclusion: The study shows the DL-based auto segmentation model for the radiotherapy of left breast cancer patients is accurate and feasible for most OARs except LAD. Proffered Paper Pooled analysis: dosimetric comparison for photons versus protons for renal stereotactic body radiotherapy Harshani M Green 1,2,3 , Orla Byrne 4 , Julia Henderson 4 , Pravesh Bhudia 5 , Vasileios Rompokos 4,6 , Andrew Gosling 6 , Callum Gillies 6 , Colin Baker 6 , Anita Mitra 7 , Daniel Henderson 8 , Alexandra Taylor 9,2 , Vincent Khoo 1,2,10 1 Uro-oncology, Royal Marsden NHS Foundation Trust, London, United Kingdom. 2 Radiotherapy, Institute of Cancer Research, London, United Kingdom. 3 Proton Beam Therapy, University College London Hospitals NHS Foundation Trust, London, United Kingdom. 4 Physics, Royal Marsden NHS Foundation Trust, London, United Kingdom. 5 Proton Beam Therapy Dosimetry, University College London Hospitals NHS Foundation Trust, London, United Kingdom. 6 Proton Beam Therapy Physics, University College London Hospitals NHS Foundation Trust, London, United Kingdom. 7 Uro-oncology, University College London Hospitals NHS Foundation Trust, London, United Kingdom. 8 Radiotherapy, Queen Elizabeth Hospital Birmingham, London, United Kingdom. 9 Radiotherapy, Royal Marsden NHS Foundation Trust, London, United Kingdom. 10 Department of Medical Imaging and Radiation Science, Monash University, VIC, Australia Purpose/Objective: There are minimal data comparing different delivery platforms for primary Renal Cell Carcinoma (RCC) SBRT. Grade 1-2% toxicity is up to 38%(1); development of patient-selection tools may aid reduced toxicity. We aimed to pool data from two SBRT planning studies: one in small RCC (<4cm)(2) and the other in large RCC (≥4cm)(3). These compared three treatment platforms across 2 institutions: C-arm linear accelerator-based volumetric modulated arc therapy (VMAT), Cyberknife (CK) and pencil-beam scanning (PBS) proton beam therapy (PBT). The primary objective was to evaluate if PBT reduces small bowel D0.1cc compared to photons. Material/Methods: Sequential anonymised cases were planned with all 3 platforms to 42Gy/Gy(RBE) in 3 fractions(4) in both studies using the same protocol. For VMAT, PTV=ITV+5mm. For CK, fiducial markers were assumed for real-time tracking, PTV=CTV+5mm. For PBS PBT, static serial organs and ITV coverage were evaluated under robustness; 5mm (geometric) and 3.5% (range) uncertainties. D95%=100% was mandated (worst-case scenario for PBT); Dmax 123- Keywords: Auto contouring, Breast cancer 1009