S2094
Physics - Image acquisition and processing
ESTRO 2026
Christie NHS Foundation Trust, Manchester, United Kingdom
Purpose/Objective: Adaptive radiotherapy (ART) relies on accurate daily imaging, but cone-beam CT (CBCT) is limited by poor soft-tissue contrast, artefacts and inconsistent Hounsfield units (HUs), restricting its use for treatment adaptation. Deep learning-based synthetic CT (sCT) offers a potential solution by improving HU accuracy and maintaining anatomical fidelity; however, clinical translation remains limited. This exploratory international survey aimed to capture professional perspectives on the benefits, barriers, and validation needed for CBCT-based sCT, to guide its safe and effective implementation in ART. Material/Methods: A cross-sectional web-based questionnaire was distributed to radiotherapy professionals worldwide. The 40-question survey included multiple-choice, Likert-scale, and open-ended items, structured across five workflow-relevant domains:(1) current CBCT use and adaptive practice,(2) awareness and experience of sCT,(3) validation requirements for image guidance, segmentation, and dosimetry,(4) Quality Assurance (QA) requirements (both for initial implementation and ongoing QA), and; (5) professional demographics.The survey targeted clinical, technical, and research experts. Quantitative responses were summarised descriptively, and qualitative free-text responses were analysed thematically to identify shared priorities, concerns, and implementation barriers. Results: Twenty-five responses from ten countries were analysed, representing medical physicists (62.5%,16), radiographers (16.7%, 4), research scientists (12.5%,3), and clinicians (8.3%, 2). Respondents highlighted strong potential for sCT (Figure 1) in supporting dose recalculation (88%) and improving workflow efficiency (66%). However, major barriers (Figure 2) included the lack of standardised validation processes (88%), challenges with system integration (72%), and insufficient transparency of the deep learning model (60%). QA was identified as crucial, with 67% supporting per-patient checks, but warnings against excessive workload were prevalent. Validation priorities differed by clinical task, with local similarity deemed most important for image guidance (80% ranked very important or above), registration accuracy for segmentation (91%), and dosimetric analysis for dose calculation (80%). Free-text responses emphasised the need for clear clinical guidance and vendor transparency to support safe adoption across the radiotherapy workflow.
Conclusion: GRE-based B0 field mapping provides reliable geometric accuracy estimation after calibration in-vivo in low-field MRI when susceptibility artifacts are moderate, as with titanium or cobalt-chrome implants. For severe distortions caused by stainless-steel, more advanced sequence models are required to incorporate sequence-dependent effects, such as pile-
up artifacts. References:
1. Keesman, R., et al., Clinical workflow for treating patients with a metallic hip prosthesis using magnetic resonance imaging-guided radiotherapy. Physics and Imaging in Radiation Oncology, 2020. 15: p. 85- 90.2. Dymerska, B., et al., Phase unwrapping with a rapid opensource minimum spanning tree algorithm (ROMEO). Magnetic Resonance in Medicine, 2021. 85(4): p. 2294-2308. Keywords: MRI, distortions, implants Towards clinical translation of synthetic CT in adaptive radiotherapy: initial insights from an international survey Chelsea Sargeant 1 , Jane Shortall 1 , Robert Chuter 2,1 , Alan McWilliam 1 1 Division of Cancer Sciences, School of Medical Sciences, Faculty of Biology, Medicine and Health, The University of Manchester, Manchester, United Kingdom. 2 Christie Medical Physics & Engineering, The Poster Discussion 4553
Made with FlippingBook - Share PDF online