3201 - A Systematic Data-Driven Process to Support Deployment of Radiation Therapy Planning Automation

Purpose/Objective(s): Radiation therapy (RT) planning automation (PA) improves dosimetric quality, planning consistency and workflow efficiency. Despite convincing evidence, scaled adoption of automation has been limited. A key deployment challenge is variability in radiation oncologists’ planning priorities. Clinical Alignment (CA) is a systematic data-driven process to elicit institutional and physician preferences and to fine-tune automation models accordingly. We present results from CA at a US community RT center.

Materials/Methods: We randomly selected 25 prostate RT plans treated during the past year. PA was performed for these. Manual refinement, standard in clinical practice, was deliberately omitted to test automation capabilities in isolation. A quality report (DVH analysis & dose distributions) was reviewed by key stakeholders, and PA models were iteratively refined.

Results: Clinical and automated plans were compared for Prostate only (n=10), Prostate and Nodes (n=10), and Prostate SABR (n=5). PA achieved clinically equivalent target coverage with improved rectal and bladder sparing in the intermediary dose range while meeting consensus constraints (see Table). PA resulted in narrower interquartile ranges (IQR).

Conclusion: CA was a valuable tool for all stakeholders during PA roll out: aligning on planning priorities, developing fine-tuned PA models, demonstrating improved planning consistency and highlighting areas for standardization. PA has been successfully deployed since CA.