S2228
Physics - Intra-fraction motion management and real-time adaptive radiotherapy
ESTRO 2026
deformations, demonstrating its potential for identifying patients that need a treatment plan adaptation without additional imaging dose, supporting low-burden online decision making. References: Boersma LJ, Sattler MGA, Maduro JH, Bijker N, Essers M, van Gestel CMJ, Klaver YLB, Petoukhova AL, Rodrigues MF, Russell NS, van der Schaaf A, Verhoeven K, van Vulpen M, Schuit E, Langendijk JA. Model-Based Selection for Proton Therapy in Breast Cancer: Development of the National Indication Protocol for Proton Therapy and First Clinical Experiences. Clin Oncol (R Coll Radiol). 2022 Apr;34(4):247-257. doi: 10.1016/j.clon.2021.12.007. Epub 2022 Jan 5. PMID: 34996684. Keywords: SGRT, dose prediction, breast surface deformation Feasibility of patient-specific motion management QA using the ZEUS MRgRT phantom for MR-guided online adaptive abdominal SBRT Gilles Weinhoffer 1,2 , Artem Napov 2 , Zelda Paquier 3 , Akos Gulyban 1 1 Medical Physics, Hôpital Universitaire de Bruxelles (HUB), Institut Jules Bordet, Brussels, Belgium. 2 Métrologie Nucléaire, Université Libre de Bruxelles, Brussels, Belgium. 3 Medical Physics, Hôpital Universitaire de Bruxelles (HUB), Institut Jules Bordet, Brussel, Belgium Purpose/Objective: The integration of MRI into radiotherapy allows real- time tumor tracking and adaptive treatments. The Unity (Elekta, Stockholm, Sweden) system includes a Comprehensive Motion Management (CMM) algorithm that predicts tumor motion 240 ms ahead to overcome latency between imaging and irradiation. However, its performance under irregular or complex breathing conditions is not fully known. This study aimed to test the feasibility of using the ZEUS MRgRT motion phantom (Sun Nuclear, Melbourne, FL, USA) for patient-specific motion management quality assurance (QA). The goal was to evaluate: 1) Digital Poster 2113 Transforming patient specific motion to be used with the phantom, 2) mimic patient movement and to test the CMM algorithm, 3) identify the algorithm’s limits under clinically challenging respiratory patterns such as irregular breathing, baseline drifts, and anatomical deformation. Material/Methods: Respiratory data from two abdominal cancer patients (seven sessions in total) were extracted from Elekta Unity Auditlog files. A Python-based pipeline was developed to convert these data into motion files
from twenty-eight patients were used to model the relationship between calculated surface deformation and (1) the CTV D98% and (2) NTCP for ACE. Surface deformation assessment and the resulting prediction models were implemented in a novel tool: Dose Impact Prediction from Surface Imaging (DIPSI, Figure 1). DIPSI was used to determine the need for plan adaptation based on the clinical thresholds CTV D98% >94% prescribed dose or Δ NTCP > 2.5% for the remaining eighteen patients (validation set). Finally, 29 clinical SGRT images acquired during repeat CT (rCT) were evaluated using DIPSI to compare predicted versus clinically indicated plan adaptations.
Results: Figure 2 shows for the validation set the CTV D98% and ACE NTCP predictions from the surface deformations as a function of true values, with R ² values of 0.860 and 0.799, respectively. DIPSI identified required plan adaptations with 99.9% sensitivity and 98.2% specificity. When applied to clinical SGRT images, DIPSI achieved 100% agreement with rCT- based adaptation decisions. The DIPSI workflow takes seconds, supporting integration into online adaptive treatment workflows.
Conclusion: DIPSI enables fast and reliable SGRT-based assessment of the dosimetric impact of breast surface
Made with FlippingBook - Share PDF online