ESTRO 2025 - Abstract Book

S4316

RTT - Treatment planning, OAR and target definitions

ESTRO 2025

period: one year before (pre-AI) and one year after AI implementation (post-AI). Descriptive statistics summarized the data for all patients and within Abdomen, Breast, H&N, Thorax, Prostate, and Pelvis cohorts.

Results: A total of 2376 patients were included in the pre-AI cohort, and 1830 patients in the post-AI cohort. Overall, AIseg reduced the median RT-led OAR contouring task time from 3.4 hours to 1.6 hours, a 54% decrease (the mean RT-led contouring time reduced from 6.1 hours to 4.5 hours, a 25% decrease). Outlier analysis also showed reduced times for cases above the 75th percentile, though extreme outliers (95th+ percentile) showed no significant improvement. Post-AI implementation, median RT-led contouring task times decreased variably across cohorts: Thorax by 68.1% (n=537/408; p<0.01), H&N by 57.7% (n=457/316; p<0.01), Abdomen by 56.8% (n=68/57; p<0.04), Prostate & Pelvis by 30.5% (n=746/569; p<0.01). Breast by 11.1% (n=149/190; p=0.09), and all other both 43.5% (n=419/290; p<0.01). Although the time spent actively contouring was reduced, the interval between cases becoming available for contouring and subsequent completion of the task, or the overall workflow time from planning-CT to treatment commencement, did not reduce for any cohort.