S2370
Physics - Quality assurance and auditing
ESTRO 2026
Stockholm, Sweden) using six clinically representative beam sets created on the rCT, copied to each sCT and recalculated without re-optimization: Beam-Set (BS) 1, four-field coplanar three-dimensional conformal radiotherapy (3D-CRT); BS2-BS5, single coplanar 3D- CRT fields from anterior, posterior, left-lateral, right- lateral; BS6, single-arc Volumetric Modulated Arc Therapy. Dose-Volume-Histrogram (DVH) metrics included D98 and Dmean. Acceptability followed German national stereotactic criteria ( Δ D98, Δ Dmean ≤ 3%) [4].
reduced QA execution time compared to phantom- based verification, offering a tangible benefit in clinical efficiency. Conclusion: LINACWatch® provides a robust and efficient logfile- based alternative for radiotherapy QA, capable of real- time error detection with high accuracy. Its integration into daily workflows supports a paradigm shift toward automated, data-driven QA that enhances both safety and productivity in modern radiotherapy practice. Keywords: Quality Assurance (QA), LINACWatch, VMAT Benchmarking head phantoms for synthetic computed tomography quality assurance in magnetic resonance-only radiotherapy Bernd-Niklas Axer 1,2 , Johann Brand 1 , Lucas Pieper 1 , Florian Putz 1 , Johanna Lott 3 , Christoph Dickmann 3 , Stefanie Corradini 1 , Rainer Fietkau 1 , Christoph Bert 1 , Juliane Szkitsak 1 1 Department of Radiation Oncology, University Hospital Erlangen, Erlangen, Germany. 2 Innovation Strategy and Technology, Siemens Healthineers AG, Forchheim, Germany. 3 Cancer Therapy Imaging, Siemens Healthineers AG, Forchheim, Germany Proffered Paper 2865 Purpose/Objective: Using magnetic resonance(MR)-only workflows can avoid computed tomography (CT)-MR registration errors and the resulting dosimetric uncertainties [1], yet standardized synthetic CT (sCT) Quality Assurance (QA) is lacking and dedicated head phantoms are scarce [2, 3]. We benchmark commercially available head phantoms regarding anthropomorphic realism to identify a phantom for robust, routine sCT QA in radiotherapy. Material/Methods: Seventeen head phantoms were screened. Nine met predefined inclusion criteria (e.g., long-term durability, gross head-shape similarity) and were imaged on a Siemens Healthineers MAGNETOM Sola 1.5T (Figure 1). For each, a T1-VIBE-Dixon-sequence needed for sCT generation was acquired. sCTs were generated in syngo.via RT Image Suite (Siemens Healthineers AG, Forchheim, Germany) using a 2D slice-wise commercial model (2DsCT,VB60) and a 3D neural- network prototype applied to the full volume (3DsCT,VC10). A planning CT acquired on a Siemens Healthineers SOMATOM go.Open Pro served as gold standard and reference (rCT). Similarity metrics included Hounsfield Units (HU)-error (mean±SD) for bone/soft tissue and geometric evaluation via Dice similarity coefficient (DSC) and 95th-percentile Hausdorff distance (HD95). Dosimetry was evaluated in RayStation (v12A, RaySearch Laboratories,
Results: Bone metrics are reported only for phantoms with skull structure (E-I). For bone-segmentation geometry, the best phantom was F (HD95=4.98/6.09mm; DSC=0.58/0.64 for 2D/3D), while the worst was G (HD95=20.53/23.54mm; DSC=0.23/0.23). Bone HU matched rCT best in H (sCT=914±471HU vs. rCT=940±270HU) and worst in E (sCT=862±453HU vs. rCT=474±132HU); soft-tissue HU agreed best in H (sCT=31±35HU vs. rCT=31±29HU) and worst in D (sCT=23±27HU vs. rCT= − 12±15HU). Beam-set-averaged DVH deviations were smallest in H ( Δ Dmean= − 0.44%, Δ D98= − 0.38%) and largest in F ( Δ Dmean=5.77%, Δ D98=4.85%) (Figure 2). Stereotactic thresholds were met by 4/9 (2DsCT) and 3/9 (3DsCT) phantoms for Δ Dmean, and by 5/9 (2DsCT) and 4/9 (3DsCT) for Δ D98. Phantom performance was similar for 2DsCT and 3DsCT, with only minor rank shifts.
Made with FlippingBook - Share PDF online