S2157
Physics - Inter-fraction motion management and daily adaptive radiotherapy
ESTRO 2026
Poster Discussion 3492 Feasibility of direct-to-treatment ultra-
hypofractionated whole-breast VMAT using AI- assisted CBCT-guided adaptive radiotherapy Shanshan Tang, Justin Visak, Tingliang Zhuang, Chang- Shiun Lin, Chien-Yi Liao, Mona Arbab, Sean Domal, Narine Wandrey, Cynthia Tye, Xingzhe Li, Prasanna Alluri, Asal Rahimi, Mu-Han Lin, David Parsons Radiation Oncology, UT Southwestern Medical Center, Dallas, USA Purpose/Objective: Direct-to-treatment radiotherapy eliminates simulation CT and pre-treatment planning, reducing the time from consultation to first fraction from several weeks to just hours. This approach can significantly improve patient experience by minimizing delays, reducing anxiety, and streamlining care. We evaluated the feasibility of delivering ultra- hypofractionated whole-breast radiotherapy (Fast Forward1) using CBCT-guided ART with VMAT, incorporating HU-corrected CBCT, AI-assisted segmentation, and a streamlined, reference-plan–less approach for real-time plan adaptation. Material/Methods: We retrospectively analyzed 22 breast cancer patients (11 left-sided, 11 right-sided) previously treated with CBCT-guided ART using VMAT. For each laterality, one patient served as a template, and the remaining 10 were test cases. The template’s planning CT, contours, and reference VMAT plan were imported into the treatment planning system to create a generalized placeholder plan. During online ART, patient-specific CBCTs were acquired, and an AI engine automatically segmented OARs, lungs, heart, stomach, spinal canal, contralateral breast, and the target (ipsilateral breast). During ART emulation, patient-specific CBCTs were utilized, and the placeholder VMAT plan was adapted directly on the CBCT image set for dose computation.Planning objectives included PTV coverage (V26Gy ≥ 90%), hotspot control (V105%<7%, V107%<2%, Dmax<110%), and hard OAR constraints: heart (V1.5Gy<50%, V3Gy<10%, V7Gy<5%) and ipsilateral lung (V8Gy<15%). Other OARs followed ALARA principles. AI-generated PTVs were compared to physician contours using the Dice similarity coefficient (DSC).
Conclusion: Iris CBCT images show substantially improved quality for reliable dose calculation and auto-segmentation. Furthermore, the implemented IMAC method effectively reduces motion streak artifacts, leading to improvements in contouring accuracy for affected cases. This comprehensive advancement, providing both quantitative accuracy and artifact suppression, facilitates the use of CBCT for robust online and offline adaptive radiotherapy on conventional C-arm Linacs. References: 1- Maier J et al. Real-time scatter estimation for medical CT using the deep scatter estimation: Method and robustness analysis with respect to different anatomies, dose levels, tube voltages, and data truncation. Med Phys. 2019;46(1):238-249. 2- Mason JH et al. Polyquant CT: direct electron and mass density reconstruction from a single polyenergetic source. Phys Med Biol. 2017;62(22):8739-8762.3- Rudin LI, and Osher S. Total variation based image restoration with free local constraints. Proceedings - International Conference on Image Processing, ICIP. 1994;1:31-35.4- Urago Y et al. Evaluation of auto-segmentation accuracy of cloud-based artificial intelligence and atlas-based models. Radiation Oncology. 2021;16(1):175. Keywords: CBCT, DL reconstruction, Motion Artifact reduction
Made with FlippingBook - Share PDF online