S2095
Physics - Image acquisition and processing
ESTRO 2026
Proffered Paper 4775 Quantitative MRI-Based Characterization of Gravity-Induced Pelvic Soft-Tissue Deformation for Effective Upright Radiotherapy Lola Vegeas 1,2 , Vincent Lepetit 3 , Niek Schroeder 4 , Tracy Underwood 5 , Nikos Paragios 6,7 , Eric Deutsch 8 , Vincent Grégoire 9 1 R&D Artificial Intelligence, Therapanacea, Paris, France. 2 Joint Collaboration: Gustave Roussy INSERM 1030 / Department of Radiation Oncology, Université Paris-Saclay / Centre Léon Bérard, Villejuif / Lyon, France. 3 Ecole des Ponts, Université Gustave Eiffel, CNRS, Paris, France. 4 Medical Physics, Leo Cancer Care, Crawley, United Kingdom. 5 Radiation Oncology, Leo Cancer Care, Crawley, United Kingdom. 6 CEO, Therapanacea, Paris, France. 7 Centrale Supélec, Université Paris-Saclay, Paris, France. 8 Gustave Roussy INSERM 1030, Université Paris-Saclay, Villejuif, France. 9 Radiation Oncology, Centre Léon Bérard, Lyon, France Purpose/Objective: Accurate assessment of posture-dependent anatomical variation is essential for maintaining geometric and dosimetric precision in upright radiotherapy [1]. In the pelvis, gravitational forces acting on compliant soft tissues and mobile organs lead to complex deformations that can alter target positioning and dose distributions. This study aims to quantify gravity-induced soft-tissue deformation using a robust, multi-metric deformable registration framework applied to paired supine and upright MRI scans, and to analyze organ-specific deformation patterns relevant to upright treatment planning. Material/Methods: Paired 3D MRI datasets were acquired from ten healthy volunteers in supine and upright positions. A multi-metric, AI-powered deformable registration algorithm, combining voxel- and statistical-based similarity metrics with biomechanical regularization, was employed to achieve diffeomorphic mappings between postures. This approach ensured anatomically consistent correspondence while capturing both global displacements and local shape changes. Deep learning-based segmentation of the prostate, bladder, rectum, small bowel, and colon enabled organ-wise quantification of deformation magnitude, direction, and local strain. Registration accuracy was evaluated via Dice similarity coefficients to confirm reliable structural correspondence across postures. Results: The proposed framework achieved high anatomical fidelity, with mean Dice coefficients of 0.785 across major organs. Organ-specific analyses revealed distinct soft-tissue deformation behaviours. The prostate exhibited the highest local strain (mean
Conclusion: This survey provides initial multi-national insights into clinical and technical perspectives on CBCT-based sCT. The findings highlight both enthusiasm for clinical adoption and clear requirements for evaluation standards, model transparency, and feasible QA frameworks. Participants underscored the importance of establishing consensus guidelines to bridge the gap between algorithmic performance and clinical trust. These insights contribute to defining a clinically informed roadmap for the safe and effective translation of synthetic CT into adaptive radiotherapy practice. Keywords: synthetic CT, adaptive radiotherapy, validation
Made with FlippingBook - Share PDF online