Development of Simulation-Omitted Prostate (SO-PRO) Radiotherapy Workflow for MR-Linac
Abstract
Purpose
Simulation-omitted prostate SBRT (SO-PRO) has potential to reduce treatment delays and expand access to care but presents unique challenges for SBRT workflows that require careful management of target and organ-at-risk (OAR) dose. A major barrier is diagnostic MRI—preferred for patient-specific anatomical information—does not include rectal spacer, resulting in substantial anatomical mismatch for spacer-affected organs and limiting reliability of deformed reference contours. This study evaluates the development of an MR-Linac–based SO-PRO workflow, integrating AI-assisted contouring to enable safe, efficient online adaptation.
Methods
A SO-PRO workflow was implemented where physicians identified candidate prostate SBRT patients for simulation-omitted treatment. Two MRI datasets were used for initial testing: pre-spacer diagnostic MRI to generate primary SO-PRO plan, and post-spacer MRI simulation as backup reference. Plans were created using identical prescription and constraints (45/40 Gy SIB in 5 fractions). To mitigate anatomical discrepancies introduced by the absence of spacer on diagnostic MRI, AI-based rectum and bladder contouring was integrated. During treatment, SO-PRO plans followed the standard MR-Linac online adaptive workflow with physician review and approval prior to delivery.
Results
Four patients were treated using SO-PRO, all successfully delivered using plans generated from diagnostic MRI. All adaptive plans met target coverage and OAR constraints without workflow failure. AI-generated contours demonstrated improved agreement for spacer-affected organs compared with deformed reference contours, with Dice similarity coefficients of 0.94 for rectum and 0.95 for bladder. Total on-couch time for patients was within one-hour, with extended sessions attributable to bladder filling and repeat adaptation rather than limitations of SO-PRO workflow.
Conclusion
A SO-PRO SBRT workflow incorporating diagnostic MRI and AI-assisted contouring is operationally feasible on MR-Linac while maintaining plan quality and adaptive efficiency. This framework is scalable to future patient-specific focal boost strategies, positioning SO-PRO as a foundation for personalized prostate SBRT and broader simulation-omitted adaptive workflows.