ESTRO 2025 - Abstract Book

S2559

Physics - Autosegmentation

ESTRO 2025

4277

Digital Poster Evaluating the organizational impact of auto-segmentation software on clinical workflow: a one-year review Jean-Luc Ley 1,2 , Étienne Delisle 1 , Danis Blais 1 1 Département de Radio-oncologie, Centre hospitalier de l'Université de Montréal (CHUM), Montréal, Canada. 2 Axe Imagerie et Ingénierie, Centre de recherche du Centre hospitalier de l'Université de Montréal (CRCHUM), Montréal, Canada Purpose/Objective: In 2023. our clinic implemented an auto-segmentation solution (AI-RAD Companion Organs RT VA50, Siemens Healthineers) to improve the efficiency of our departmental human resources and streamline clinical workflows. A multi-metric evaluation approach—including geometric, qualitative, and time-efficiency— [1] was employed to assess the solution’s organizational impact over a one-year period. By integrating and analyzing all the collected data, we aim to determine whether the solution has had a measurable organizational impact. Material/Methods: The auto-segmented organ-at-risk (OAR) contours were generated for all patients (approximately 2,700 patients after a year of clinical use) and imported into the treatment planning system (TPS, Eclipse v16.1, Varian Medical Systems). Radiation oncologists subsequently reviewed and adjusted these contours as needed. A record of the original contours, alongside the modified versions, was maintained within the TPS. To assess quality, a Likert-scale survey integrated into the record-and-verify system (R&V, MOSAIQ v2.64, Elekta Inc.) allowed clinicians to provide feedback on the overall quality of the contours. Workflow efficiency was evaluated through an analysis of task timestamps extracted from the R&V, enabling a comparison of pre- and post-implementation workflows. Results: The Likert-scale survey, with over 1,100 responses (response rate: 40.9%), showed that nearly 90% of cases required only minor or no corrections. Figure 1 details these subjective assessments by anatomical site, revealing overall high satisfaction with site-specific variations. Gynecological and gastrointestinal sites had the highest proportion of clinically acceptable OARs, while the head and neck region proved the most challenging, with the highest proportion of "Must Redo".