S2138
Physics - Inter-fraction motion management and daily adaptive radiotherapy
ESTRO 2026
deviation from the planned dose. Interpretation of rectal dose relative to constraints was limited due to contouring-guideline differences between clinical contours and daily auto-contouring.For the bladder, 57.90% of fractions delivered higher D1% values than planned, and seven patients presented ≥ 3 consecutive sessions with >5% deviation from the planned dose. Moreover, these seven patients also did not comply with the dose constraint (D1% < 80 Gy).
Analysis Toolkit for Medical Imaging in Python. J Open Source Softw. 2023;8(86):5374. Keywords: image-guidance, deep-learning, childhood cancer
Digital Poster Highlight 2476 Implementation of an AI-Driven Workflow for Daily Dose Reconstruction in Prostate Cancer Radiotherapy Jessica Prunaretty, Tom Baudouin, Olivier Riou, David Azria, Pascal Fenoglietto Radiotherapy, Institut du Cancer de Montpellier, Montpellier, France Purpose/Objective: This study evaluated the daily delivered dose for patients enrolled in the experimental arm of the RCMI- GI prostate cancer trial using a fully automated AI- based software, Adaptbox (Therapanacea). The objective was to assess target coverage and organ-at- risk (OAR) exposure under reduced PTV margins. Material/Methods: Twenty patients treated within the clinical trial were included. All received 80 Gy in 40 fractions to PTV2 and 56 Gy simultaneously to PTV1 using two-arc VMAT on a TrueBeam STx. Daily setup was performed with CBCT imaging, and intra-fraction prostate motion was monitored using real-time tracking (Calypso system, Varian).For each treatment fraction, the CBCT was imported into Adaptbox (v2.3.2). A synthetic CT (sCT) was generated from the CBCT using a deep learning– based algorithm. OARs were automatically segmented, while target structures were propagated from the planning CT (pCT) by rigid registration. Dose calculation was performed using Adaptbox’s proprietary collapse-cone algorithm.To reduce discrepancies previously identified between Adaptbox and clinical contours [1], the Adaptbox rectal auto- contouring model was also applied to the pCT. A normalization factor aligning Adaptbox dose calculations with Eclipse TPS doses on the pCT was determined for each patient and for each dose metric, and subsequently applied to all fractions. RCMI-GI trial dose parameters were extracted for each session and compared with the planned values. Results: All 800 delivered fractions were analyzed. Target coverage remained consistent with planning for all patients, with a maximum deviation of 0.2% for both PTV2 and PTV1.For the rectum, most fractions delivered higher doses than planned: 81.04%, 35.78%, and 77.24% of sessions exceeded the planned dose for D0.1%, V70Gy, and V76Gy, respectively (Figure 1). Three patients (V70Gy) and one patient (V76Gy) experienced ≥ 3 consecutive fractions with >5%
Conclusion: This AI-based sCT workflow enables reliable daily dose reconstruction and highlights relevant deviations in OAR exposure that could trigger adaptive interventions. Harmonization of contouring practices remains essential to ensure consistent interpretation of daily dose results. References: [1] Prunaretty et al, Evaluation of the contour generation and dose calculation algorithm of a AI- powered offline adaptive software for prostate cancer, ESTRO 2026 submission Keywords: Adaptbox, prostate, daily dose MOST SMART Prostate: A streamlined MRI-only workflow from 3T simulation to 0.35T MR-Linac SBRT Marvin Kinz 1,2 , Jennifer W Campbell 1 , Cassandra L. Bullens 1 , Jürgen Hesser 3,4 , Evangelia Kaza 1 , Yue-Houng Hu 1 , Dianne Ferguson 1 , Shu-Hui Hsu 1 , Zhaohui Han 1 , Jeremy Bredfeldt 1 , Jonathan E. Leeman 1 , Atchar Sudhyadhom 1 , Kamal Singhrao 1 1 Department of Radiation Oncology, Brigham and Women's Hospital, Dana-Farber Cancer Institute, Harvard Medical School, Boston, USA. 2 Department of Physics and Astronomy, Interdisciplinary Center for Digital Poster Highlight 2519
Made with FlippingBook - Share PDF online