ESTRO 2025 - Abstract Book

S4347

RTT - Treatment planning, OAR and target definitions

ESTRO 2025

selected, Limbus Contour (LB) and MVision Contour+ (MV), for independent RO evaluation based only on contour delineation quality. The aim of this work is to determine which software best suits our department. Material/Methods: Generated OARs were divided into five anatomical areas (Head&Neck, Thorax, Breast, Abdomen and Pelvis) and blindly presented for evaluation to ten ROs from our department. Contours were compared using a 3-point Likert scale: Not used , Incomplete / Major edition , Minor edition / Unedited . Due to demo limitations in number of scans, a single delimitation per CT scan was performed, prohibiting direct comparison between LB and MV per case, and yielding different number of cases per software, anatomical area and RO. Results: OAR comparison between both software packages by anatomical area (Figure 1) showed the following relevant results: 1. LB generally performed better for Head&Neck and Thorax: Unedited/Minor edition rate are 71.2% and 86.5%, respectively; 2. MV performed better for Breast, Abdomen and Pelvis with a best score of 82.3% for the Abdomen. A global analysis of all the OARs evaluated (Figure 2) shows that the results are very similar for both softwares, with LB and MV showing 64.1% and 67.5% respectively for the Unedited/Minor edition classification. AI generated contours were Not used at all in 12.6% of cases for LB and 10.0% for MV. The variability in the evaluation criteria between all of RO could impact the results.