ESTRO 2025 - Abstract Book

S3081

Physics - Inter-fraction motion management and offline adaptive radiotherapy

ESTRO 2025

Results: AI contours on low-dose CBCT were consistently smaller than those delineated by ROs, with differences of -10.4% ± 6% for parotid glands and -59% ± 19% for submandibular glands, likely reflecting limitations of applying CT-trained models to older CBCT systems. While variable on a fraction-to-fraction basis, AISeg correctly detected significant anatomical changes (>20%). Parotid volume change showed the strongest correlation with weight loss, with a 20% change by fraction 15 identifying 65% of patients who subsequently experienced >10% weight loss, though with a 25% false positive rate. Applying the 1mm surface overlap metric with a 60% threshold improved true positive rates to 77% with a false positive rate of 32%. Moderate to strong correlations between final structure volume and earlier treatment periods were found. Parotid correlations increased from moderate (0.48–0.57) over the first 5 days to strong (0.76–0.88) by day 15. Although not used clinically for these cases, XB nodal levels showed strong correlations by day 15 (0.74–0.77). Submandibular glands showed weaker correlations, increasing from 0.21–0.32 at 5 days to 0.51–0.60 at 15 days. Conclusion: Early changes in raw AISeg volumes may be sensitive indicators of significant anatomical variation, supporting their use as low-workload predictive tools for adaptive radiotherapy. As AI models for CBCT emerge, validation across diverse CBCT systems will be essential. Digital Poster Anatomical robustness of a shuttle-based workflow for daily online MRI-guided particle therapy Friderike K. Longarino 1,2,3 , Una Maguire 1,2,4 , Cedric Beyer 2,3 , Rita Pestana 2,3 , Sebastian Regnery 2,3 , Jürgen Debus 2,3,5 , Sebastian Klüter 2,3 , Katharina Seidensaal 2,3 , Julia Bauer 2,3 1 Clinical Cooperation Unit Translational Radiation Oncology, German Cancer Research Center (DKFZ), Heidelberg, Germany. 2 Department of Radiation Oncology and Heidelberg Ion Beam Therapy Center, Heidelberg University Hospital, Heidelberg, Germany. 3 Heidelberg Institute for Radiation Oncology (HIRO), National Center for Radiation Research in Oncology (NCRO), Heidelberg, Germany. 4 School of Physics, University College Dublin, Dublin, Ireland. 5 Clinical Cooperation Unit Radiation Oncology, German Cancer Research Center (DKFZ), Heidelberg, Germany Purpose/Objective: Online adaptive particle therapy makes it possible to consider anatomical changes during treatment, which may lead to improved treatment outcomes [1,2]. A shuttle-based daily quasi-online MRI-guided strategy, where the patient remains in the treatment position during transfer between MRI and irradiation, aims to achieve this end while minimizing workflow disruption. This study investigated potential liver displacement and deformation while using a patient-transfer shuttle system with our current carbon ion SBRT (stereotactic body radiotherapy) immobilization set-up, with the goal of applying the workflow for online MRI-guided particle therapy. Material/Methods: Fourteen healthy volunteers (age: 24–61 years) underwent a series of MRI scans in a 1.5 T MRI scanner using a respiratory-triggered 3D T2 TSE sequence, interspersed with intra-hospital transport simulations. The study comprised (Figure 1): volunteer positioning on a vacuum mattress fixed to a transfer table with abdominal compression, MRI 1 (baseline MRI), a ten-minute waiting period as a time control phase, MRI 2, a short transport phase with a trolley, MRI 3, a longer transport phase with a trolley, and MRI 4. In each MRI set, the liver and the external outline for the region encompassing the liver were contoured. The Dice similarity coefficient (DSC) and the mean distance to agreement (MDA) were calculated for the accumulative shifts throughout the study and for shifts between consecutive image sets. Keywords: AI segmentation, CBCT, anatomical indicators 639