S1134
Clinical - Urology
ESTRO 2026
Conclusion: In a five-fraction prostate SBRT cohort, 2-year clinician- assessed GU/GItoxicity was low and consistent with PACE-B. PCA-guided DVH features—rectum V13% andrectal-wall high-dose (PR_37.5) for GI, plus bladder- wall high-dose metrics for GU—showuseful discrimination, with bladder predictors chiefly providing high NPV.These findings support integrating PCA-selected DVH features and calibrated rule-outcut- offs alongside guideline constraints in treatment planning and for parsimoniousmultivariable models. References: Tree AC, Ostler P, van der Voet H, Chu W, Loblaw A, Ford D, et al.Intensity-modulated radiotherapy versus stereotactic body radiotherapy for prostatecancer (PACE-B): 2-year toxicity results from an open-label, randomised, phase 3,non-inferiority trial. Lancet Oncol. 2022;23(10):1308-1320.doi:10.1016/S1470- 2045(22)00517-4. Keywords: prostate, SBRT, DVH, PCA Geometric evaluation of a deep learning method for segmentation of urinary OARs on magnetic resonance imaging, for prostate cancer radiotherapy Jennifer Le Guevelou 1 , Miguel Castro 2 , Blanche Texier 2 , Anais Barateau 2 , Romane-Alizé Martin 2 , Caroline Lafond 2 , Igor Bessieres 3 , Jean-Claude Nunes 2 , Renaud De Crevoisier 2 , Oscar Acosta 2 1 radiation oncology, CHRU Tours, Tours, France. 2 IMPACT, LTSI, Rennes, France. 3 physics, Centre Georges François Leclerc, Dijon, France Purpose/Objective: Purpose: While urinary organs at risk (OARs) such as the intraprostatic urethra and the bladder trigone are increasingly recognized as associated with severe genitourinary toxicity, their delineation in clinical practice is time consuming and probably associated with a large interobserver variability. The aim of this study was to propose a magnetic resonance (MR) deep learning segmentation of urinary OARs for prostate cancer (PCa) radiotherapy (RT), based on a validated atlas. Material/Methods: Material and methods: In this multicentric study, a convolutional neural network (CNN) for image segmentation (nnU-Net) was trained and validated on three image datasets. Two datasets came fromMR- linac devices (Unity®, Elekta et MRIdian®, Viewray), and one dataset came from the PROSTATEx database (MAGNETOM® Trio and Skyra, Siemens). Evaluation of the deep learning segmentation was performed using dice score coefficients (DSC), surface distance (SD) and Digital Poster 274
Hausdorff distance. Results:
Results: A total of 265 MRI were analyzed. The mean DSC for all urinary structures was 0.88. The automatic segmentation model proved to be effective in the segmentation of the target volume and large OARs such as the bladder and the rectum (mean DSC ranging between 0.87 to 0.95). Regarding urinary OARs, the mean DSC ranged between 0.50 to 0.68. The Hausdorff distance ranged between 5.5mm – 27.6mm for urinary OARs, highlighting local mismatches caused by large anatomical variations between patients. However, the SD ranged between 1mm and 1.4mm for urinary OARs, highlighting an overall good surface correlation for all organs. An example is provided in Figure 1. Conclusion: Conclusion: This multicentric study is the first to propose a nnU-Net deep learning model for the delineation of urinary OARs, that can be applied to various image dataset. Further work is needed to assess the dosimetric impact of such variations, in various clinical scenarios. References: Le Guévelou, J. et al. Urinary Organs at Risk for Prostate Cancer External Beam Radiation Therapy: Contouring Guidelines on Behalf of the Francophone Group of Urological Radiation Therapy. Pract. Radiat. Oncol. S1879-8500(24)00145–0 (2024) doi:10.1016/j.prro.2024.05.009. Keywords: segmentation, urethra, urinary organs at risk
Made with FlippingBook - Share PDF online