ESTRO 2025 - Abstract Book

S4083

RTT - Patient care, preparation, immobilisation and IGRT verification protocols

ESTRO 2025

Conclusion: Our results demonstrate promising potential for generating synthetic CT (synCT) from external contours alone. Further studies are planned to refine this approach and evaluate its performance in a clinical setting.

Keywords: Synthetic CT, AI, Diffusion model

References: [1] Sohl-Dickstein, J., Weiss, E. A., Maheswaranathan, N., & Ganguli, S. (2015). Deep unsupervised learning using nonequilibrium thermodynamics. Proceedings of the 32nd International Conference on Machine Learning (ICML), 2256-2265.

2039

Digital Poster Intrafractional patient movement during online adaptive radiotherapy for rectal cancer patients Sophie L. van Rossen, Judith H. Sluijter, Claudia S.E.W. Schuurhuizen, Maarten L.P. Dirkx, Joost J. Nuyttens Radiotherapy, Erasmus MC, Rotterdam, Netherlands Purpose/Objective: In online adaptive radiotherapy (OART), only intrafractional movement needs to be considered, as the clinical target volume (CTV) is daily delineated using cone-beam CT (CBCT) imaging, thereby eliminating interfraction movement. In this study, we investigated intrafraction patient movement during OART for rectal cancer patients. Material/Methods: This retrospective study included twenty rectal cancer patients from the TRIPOS-II study (IRB protocol MEC-2023 0445). Nine patients received 5x5 Gy and eleven patients received 25x2 Gy with concurrent chemotherapy. Patients were treated using margins of 5 mm in the left-right and posterior directions, and 8 mm in the superior-inferior and anterior directions around the CTV to obtain the planning target volume (PTV). The online adaptive workflow started with the acquisition of a pre-treatment CBCT scan to define the daily patient anatomy and generate an adapted CTV and adaptive treatment plan. Prior to treatment delivery, a second CBCT scan was acquired to evaluate possible changes in patient position and anatomy and to correct for this with a couch shift. At the end of each treatment fraction, a third, post-treatment CBCT scan was acquired. To evaluate intrafractional patient movement, the post-treatment CBCT scan was registered to the pre-treatment CBCT scan based on bony anatomy in AriaEclipse (version 17.00.00), accounting for both translational and rotational movement. Systematic (Σ) and random errors (σ) were calculated for all directions using the error parameters from Brand et al. [1]. The van Herk margin recipe was used to calculate a CTV-to-PTV margin, based on translations alone [2]. The difference in fractionation schedule was excluded by deriving an effective systematic and random error [3], resulting in an effective van Herk margin for either 5 or 25 fractions. Results: A total of 310 treatment fractions were included in this study. Ten fractions were excluded due to the absence of the post-treatment CBCT scan. The average time between the pre-treatment and post-treatment CBCT scans was 26.0 ± 5.4 minutes (1 SD). For 85% of fractions, the 3D intrafraction movement was ≤1 mm; for 77% of fractions, the rotations were ≤1° in all directions. Based on the translations, a CTV-to-PTV margin of only 1.1-1.2 mm is sufficient to account for intrafractional variation in the bony anatomy (Table 1).