Automated Low-Dose Radiotherapy Planning for Knee Osteoarthritis Using a Centralized, Parameterized Planning Framework.
Abstract
Purpose
Automated treatment planning can reduce manual workload and improve consistency; however, supporting new clinical indications often requires development of dedicated tools rather than reuse of existing workflows. This study describes the creation of a centralized, parameterized external beam radiotherapy planning framework to support automated low-dose radiotherapy (LDRT) for knee osteoarthritis (OA) as an initial minimum viable product (MVP).
Methods
A centralized automation framework was designed and implemented to support end-to-end external beam planning through modular tools capable of accepting dynamic clinical parameters. Knee OA was selected as the MVP indication to define required inputs, field geometry logic, and failure modes. Using this framework, twenty-three retrospective knee OA cases were planned for both TrueBeam and Halcyon systems, with automated execution of target and structure generation, standardized AP/PA field setup, beam and collimator angle selection, couch placement, density overrides, dose calculation, beam weighting, and plan normalization. Complete sim CT–to–plan generation was achieved in under 15 minutes per case. Standardized field geometry was encoded within the framework, with AP/PA fields centered at the tibiofemoral joint space and extending 8 cm superior–inferior from isocenter and 3 cm medial–lateral from bony anatomy, consistent with published knee OA LDRT methodologies. Automated plans were reviewed alongside clinically approved manual plans as a secondary qualitative validation step.
Results
The framework successfully generated clinically reviewable knee OA treatment plans across all cases. Qualitative review demonstrated acceptable joint coverage, hotspot control, and field setup. Observed failures were primarily attributable to upstream setup inputs rather than deficiencies in the automation logic.
Conclusion
This work demonstrates the feasibility of building a centralized automation framework to support new clinical site indications, with knee OA LDRT serving as an effective MVP. Purpose-built automation can reduce planning time, standardize plan setup, and establish a scalable foundation for broader automated planning deployment.