ESTRO 2025 - Abstract Book

S4232

RTT - Education, training, advanced practice and role developments

ESTRO 2025

2167

Proffered Paper Automated Feedback for Delineation: A Path for Students to Outperform RTTs? Deniece Washington 1 , Moniek van Klink - de Goeij 2 , Daniel Tabrizian 3 , Harmen Bijwaard 2,4 , Feline Schoneveld 5 , Jelle Scheurleer 2 1 School of Allied Health Professions, Fontys University of Applied Sciences, Eindhoven, Netherlands. 2 Faculty of Health, Sports and Welfare, Inholland University of Applied Sciences, Haarlem, Netherlands. 3 Faculty Engineering, Design, and Technological Innovation, Inholland University of Applied Sciences, Haarlem, Netherlands. 4 Centre for Safety, National Institute for Public Health and the Environment (RIVM), Bilthoven, Netherlands. 5 Radiotherapy department, Holland Proton Therapy Centre, Delft, Netherlands Purpose/Objective: The aim of this study was to develop and validate feedback rules in Panoptes that match the contouring accuracy typically observed in clinical practice. This method ensures uniform, instructor-independent feedback, even for large groups, by assessing the quality of delineations against clinical standards. The feedback rules aim to support students in developing contouring skills required in radiotherapy. Material/Methods: This prospective, cross-sectional study included 47 experienced RTTs from 18 radiotherapy departments in the Netherlands. RTTs contoured a thoracic case involving the lungs, esophagus, heart, and spinal cord. Interobserver variation was measured by comparing deviations from a gold standard established through expert consensus. A focus group consisting of radiotherapists, medical physicists, and educators established feedback rules based on the results, which were subsequently tested and validated using the in-house developed web-based tool, Panoptes, with a group of 16 students. Based on literature, the focus group set a 75% passing grade for student performance. Results: The study revealed relatively low contouring variation among RTTs for the left lung (3.93 ± 1.39 mm), right lung (4.64 ± 1.17 mm), and spinal cord (4.63 ± 0.80 mm) at 95% of the contoured volume. However, variability was higher for the esophagus (5.03 ± 5.77 mm) and heart (27.48 ± 16.69 mm). A statistically significant negative correlation was found between variation and work experience, contouring experience, and level of education in contouring. This indicates that as experience and education increase, variation decreases. The correlation coefficients were weak, ranging from -0.30 to -0.50. The focus group set thresholds for feedback rules, designed to challenge students without expecting them to outperform experienced RTTs. Although penalties were considered, they were excluded due to concerns about their impact on student motivation. Students demonstrate smaller deviations from the gold standard, particularly for the heart (11.09 ± 4.22 mm) and esophagus (3.43 ± 1.27 mm), outperforming RTTs in several areas. These differences were statistically significant for the spinal cord, esophagus, and lungs. The set feedback rules tested on the students’ data set at 95% contoured volume demonstrate success rates of 92.3% for the esophagus, 100% for the heart, 92.9% for the spinal cord, 87.5% for the left lung, and 62.5% for the right lung resulting in a pass grade of 75% in the included group of students. Conclusion: This study demonstrates that automated, instructor-independent feedback in Panoptes can effectively guide students in achieving clinically relevant contouring accuracy. Providing consistent, quality-based feedback improves student learning and enhances readiness for professional radiotherapy practice.

Keywords: automated feedback, delineation accuracy