Workflow Efficiency in Vaginal Cuff High Dose Rate Brachytherapy Using Artificial Intelligence-Based Organ Segmentation and Multi-Channel Cylinder Modeling.

Background/

Objectives

High dose rate brachytherapy (HDR-BT) is resource-intensive. Workflow efficiency and inter-user variability remain prevalent issues in HDR-BT. To improve workflow efficiency and reduce inter-user variability, we introduced artificial intelligence (AI)-based organ contouring (AC) and applicator modeling (AM) into our clinical workflow. Here, we present results on the impact of these tools on workflow efficiency, inter-user variability, and plan quality for vaginal cuff HDR-BT cases.

Methods

260 treated fractions were included in the analysis, half of which were treated before implementing AC and AM. Five different medical physicists performed the treatment planning. Using built-in timestamps, contouring, dose planning, and total treatment planning times were recorded for each fraction. Dosimetric data, including the dose to the highest 2cc (D2cc) of bladder and rectum, and percentage volumes of the target covered by 90-200% isodose lines (V90-200) were recorded. Analysis of variance with post-hoc Tukey tests were used to determine statistical differences between treatment planners before and after implementing AC and AM.

Results

Implementing AC and AM resulted in an overall 13.7 ± 1.7-min reduction in planning time. Inter-user variability in organ and target dose metrics was reduced for most structures. Most notably, statistically significant differences in rectum D2cc and target V95 were observed between planners before, but not after, AC and AM implementation.

Conclusions

AC and AM significantly improve workflow efficiency while maintaining plan quality. Reductions in inter-user variability and standardization of workflows may facilitate user training and robust outcome assessment.