ESTRO 2024 - Abstract Book

S3988

Physics - Inter-fraction motion management and offline adaptive radiotherapy

ESTRO 2024

patients would have been identified by the pipeline for ART, indicating either an increase in false positives or increased opportunity to identify patients who required ART who did not receive it. This demonstrates the need to use the pipeline as a clinical decision-making support tool. The automated pipeline was statistically significantly (p<0.05) quicker than the manual pipeline and current clinical assessment methods (mean time 182.5 s vs. 486.9 s and 556.4 s). Deformed contour accuracy was found to be acceptable, with 96.6% of deformed CTVs considered clinically acceptable by 2 physicists, and 89.8% of rectum contours. After rectum contours were manually corrected, 0.9% of sCT outcomes changed from green to red. It was noted that these inaccuracies in the pipeline did not limit the ability of the pipeline to identify ART patients.

Conclusion:

The automated pipeline is successful at identifying patient treatments which required ART with a high level of accuracy while reducing time and resource requirements. Figure 2 suggests that the pipeline could allow improved detection of treatments that require ART, including identification at earlier time points.

Keywords: Automation, Deep-Learning, Prostate

668

Poster Discussion