ESTRO 2025 - Abstract Book

S1836

Clinical – Upper GI

ESTRO 2025

contours by the participating centres were compared to a reference contour set by the coordinating centre with support of an experienced abdominal radiologist. Agreement between submitted and reference contours was evaluated using structure volume, Dice Similarity Coefficient (DSC), Mean Distance to Agreement (MDA), and Hausdorff Distance – 95% (HD95). Results: Delineation variations were the largest for the target, central biliary tract (CBT) and cystic duct, with relative volumes in relation to the reference contours of 36-93%, 26-348% and 29-1243%, respectively (Fig. 1). DSCs of these three structures were generally poor, with wide ranges of 0.50–0.69, 0.12-0.71, and 0.00-0.49, respectively (Table 1). Poor agreements with the reference set for the target, CBT and cystic duct are further illustrated by MDA- and HD95 values. Large variations in target contours were mainly due to differences in whether or not the stent in the CBT at tumour level was included in the target. For the CBT, these variations were primarily attributed to wide discrepancies in cranial and caudal delineation boundaries. Aside from the CBT and cystic duct, a markedly better contouring consensus was attained in the remaining OARs, specifically liver (DSC=0.94 [0.93-0.94], MDA=1.8 mm [1.6-2.1], HD95=5.8 mm [4.3-6.8]) and stomach (DSC=0.92 [0.86-0.94], MDA=1.2 mm [0.7-2.9], HD95=5.5 mm [2.2 19.8]). Conclusion: Major variations in target, CBT and cystic duct delineations led the participating centres to conclude that multicentre evaluation of the contour set for each enrolled patient should be adopted. Moreover, these variations pointed out the urgency of the implementation of comprehensive contouring guidelines.