ESTRO 2025 - Abstract Book

S3581

Physics - Quality assurance and auditing

ESTRO 2025

mapping based on RIR or DIR, as part of the pre-trial quality assurance for CURE Lung, a Scandinavian thoracic reirradiation study.

Material/Methods: Seven anonymized retrospective lung cancer cases, treated with curatively-intended reirradiation near a previously irradiated region, were distributed to six Danish RT-centers. Organs-at-risk (OARs) were delineated using AI-assisted segmentation, reviewed by an oncologist, on the CT images of the previous (CT-old) and current treatment (CT-new). Four OARs (aorta, bronchial tree, esophagus, spinal cord) were contoured. RIR and DIR from CT-old to CT-new were performed for all patient cases at each center. Consensus-structures based on the transferred OARs were generated on CT-new using the STAPLE-function with a 50% agreement threshold for each OAR. Geometric differences between the transferred OARs from individual centers and the consensus-OARs were evaluated within the volume receiving ≥2Gy on CT-new. Variation relative to consensus-structures was assessed using Surface-Dice Coefficient within 3mm (SD-3mm) and Mean-Surface-Distance (MSD). Doses were mapped from CT-old to CT-new using both RIR or DIR and converted to EQD2; and DVHs from mapped doses from all six centers were extracted. Linear mixed-effect models tested the effect of registration method (RIR vs DIR, fixed effect) on both OAR alignment and DVH-values, with patient and center as random effects. Results: Centers used a variety of software systems: RayStation (1), MIM (2), or Velocity (3). DIR significantly reduced the geometrical difference between the individual center OARs and the consensus-OAR compared to RIR, with SD-3mm values closer to 1 (p≤0.001) and MSD values closer to 0cm (p≤0.001) for DIR; across all OARs and for individual organs, see Figure 1 . DVH-analysis (Figure 2) showed significantly different DVH-values for RIR versus DIR in mapped dose-distributions for aorta, esophagus, and spinal cord (p<0.05), but not for bronchial tree (p=0.752). Due to large anatomical changes, patient 7 was excluded from DVH-analyses.