ESTRO 2025 - Abstract Book

S3080

Physics - Inter-fraction motion management and offline adaptive radiotherapy

ESTRO 2025

Bladder and bowel doses were in good agreement in all fractions. 4 out of 20 fractions exhibited variation in the rectum V40Gy greater than 5%. Notably, qualitative improvements were observed in patients where anatomical changes significantly impacted dose distribution. Conclusion: Overall, HyperSight CBCT was successfully calibrated for HU, showing only minimal differences in calculated doses compared to the planning CT. Furthermore, dose calculations on HyperSight CBCT images demonstrate improved accuracy compared to sCT-based calculations; suggesting the potential for enhanced treatment precision particularly for patients with significant variation in abdominal gas at treatment. Further studies are warranted to assess clinical outcomes based on these findings quantitatively.

Keywords: Online Adaptive Radiotherapy, Ethos, HyperSight

References: 1 Glide-Hurst et al., 2021. Adaptive radiation therapy (ART) strategies. IJROPB , 109(4)

2 El-Bared et al., 2019. Dosimetric benefits of MRI-guided stereotactic RT for pancreatic cancer. PRO , 9(1) 3 Christiansen et al., 2022. Adaptive radiotherapy reduces prostate cancer treatment toxicity. Radiother. Oncol. 167 4 Byrne, et al, 2021. Varian Ethos adaptive radiotherapy for prostate cancer: Early results. J Appl Clin Med Phys , 23(1). 5 Chen et al., 2021. Synthetic CT generation via unsupervised deep learning. Phys Med Biol , 66(11) 6 Lemus et al., 2023. Air mapping errors affecting prostate CBCT-guided adaptive radiation therapy. J Appl Clin Med Phys , 24(10). 7 Computerized Imaging Reference Systems Brochure (2016) https://shorturl.at/l5xdK

601

Digital Poster Evaluating the Efficacy of Out-of-Domain CT-Trained AI Segmentation on Low-Quality CBCT in Head and Neck (H&N) Radiotherapy Ciaran Malone 1 , Samantha Ryan 1 , Jill Nicholson 1,2 , Niall O'Dwyer 1 , Matthew Fahy 1 , Orla McArdle 1 , Frances Duane 1,2 , Aisling Nolan 1 , Ruth Woods 1 , Brendan McClean 3,1 , Sinead Brennnan 1,2 1 Department of Radiation Oncology, St.Luke's Radiation Oncology Network, Dublin, Ireland. 2 Applied Radiation Therapy Trinity, Discipline of Radiation Therapy & Trinity St James’s Cancer Institute, Trinity College Dublin, Dublin, Ireland. 3 Physics, SLRON, St.Luke's Hospital, Dublin, Ireland Purpose/Objective: This study assesses the feasibility of using raw AI auto-segmentation to monitor and predict anatomical changes in H&N cancer patients during radiotherapy. The goal is to determine if AI segmentation (AISeg) offers sensitive indicators for significant anatomical changes, potentially improving treatment outcomes through early adaptive intervention. Material/Methods: A retrospective analysis was conducted on daily CBCT scans from 35 H&N cancer patients treated with 70 Gy in 35 fractions using volumetric modulated arc therapy. The MVision AI segmentation tool, trained for CT segmentation, was used to automate contouring of OAR and nodal structures within the field-of-view of low-dose head protocol CBCT scans on Varian Linacs. For each fraction, metrics such as structure volume and average surface overlap at a 1mm threshold relative to the first fraction were recorded and trended. Correlations between these metrics and weight loss were explored. The first 15 days (10 fractions) were also evaluated against the average volume over the last 5 days as a predictive tool for identifying patients likely to undergo significant anatomical change. Independent validation of segmentation accuracy was performed by a Radiation Oncologist (RO) on selected cases.