S2229
Physics - Intra-fraction motion management and real-time adaptive radiotherapy
ESTRO 2026
Conclusion: Patient-specific QA using the ZEUS MRgRT phantom is feasible and effective for MR-guided adaptive abdominal SBRT. The phantom reliably reproduces real breathing patterns and helps test motion management algorithms under realistic and extreme conditions. The Elekta Unity’s CMM shows good robustness overall, still this test can be used to further explore (or repeat) clinical scenarios, where further improvements might be necessary. Keywords: quality assurance, MRI, motion management Digital Poster Highlight 2212 Patient-specific offline intrafraction motion verification using per-fraction triggered kV images Quentin TECHER 1,2 , Noé GRANDGIRARD 2 , Saturnin SANDJONG 2 , Vincent MARCHESI 2 , Paul RETIF 1,3 1 Medical Physics Unit, CHR Metz-Thionville, Metz, France. 2 Medical Physics Department, Institut de cancérologie de Lorraine, Vandœuvre-Lès-Nancy, France. 3 CRAN, Université de Lorraine, Nancy, France Purpose/Objective: Intrafraction motion can compromise geometric precision during radiotherapy, particularly in treatments targeting the spine and sacrum where steep dose gradients are frequently employed. While triggered kV images acquired during beam delivery contain valuable positional information, these data are rarely analyzed routinely. We developed an automated offline workflow enabling patient-specific verification of intrafraction geometry using per-fraction triggered kV images and DRR-based rigid registration. Material/Methods: The workflow imports planning CT, RTSTRUCT, RTPLAN, and triggered kV images acquired during treatment. For each image, acquisition geometry (gantry angle, source-to-imager distance, pixel spacing) was extracted from DICOM headers. A digitally reconstructed radiograph (DRR) was then generated from the planning CT at the corresponding geometry using GPU-accelerated volumetric rendering. A 2D/2D rigid registration (translation-only) was performed between each triggered kV image and its matched DRR using Mattes mutual information. Displacements along horizontal and vertical detector axes (mm) and their Euclidean magnitude were computed. A standardized PDF/CSV report was automatically generated per beam and per fraction.Validation was done in two different institutions using Varian TrueBeam accelerators and included (1) anthropomorphic phantom experiments (pelvic phantom and “Tokyo” whole-body phantom) across multiple anatomical regions, (2) a retrospective cohort
compatible with the ZEUS MRgRT phantom. The selected sessions represented typical clinical scenarios including normal, irregular, baseline-shift, and error- inducing motion patterns. The conversion included time interpolation to a constant 0.1 s sampling rate, signal smoothing, and possible adjustment to the phantom’s mechanical limits. A ±3 mm gating window was applied to simulate clinical beam control. Unity data (Auditlog.csv, gating.csv) were analyzed for tracking error, irradiation efficiency, safety behavior, and latency. Additional synthetic signals with different amplitudes and baseline drifts were created to explore extreme motion scenarios. Results: The conversion process successfully generated realistic motion profiles, confirming the feasibility of the use of the phantom for reproducing complex respiratory patterns. Based on the simulated motion situation, the CMM algorithm repeated similar behavior as in clinical scenarios: maintained good tracking accuracy for regular, small-amplitude motions, but showed reduced precision during irregular or drifting signals, with latency delays occasionally exceeding the predicted 240 ms (Figure 1). Even with large deformation (“low quality factor”) the irradiation was always safely interrupted (Figure2).
Made with FlippingBook - Share PDF online