Purpose: KBP is a deep learning process that assists planners in achieving optimal radiotherapy plans. SRS for multiple brain metastases often poses a challenge to even the best treatment planners. Our goal is to determine the efficacy of an SRS KBP model compared to a commercial planning module.
Methods: A retrospective study of fifteen patients previously treated with SRS to multiple brain lesions was performed, comparing BrainLab Multiple Metastases Elements (MME) and KBP planning for a Novalis Tx. A KBP model was generated in RapidPlan (Varian) for VMAT delivery. KBP plans utilized 4-5 arcs per isocenter; MME plans ranged from 5 to 38 conformal arcs. Arcs were manually determined by the planner for all KBP plans, whereas MME automated this process. Line optimization was utilized in KBP planning to reduce low dose spray. Ranges for prescription, number of lesions, and PTV volume were 15-24Gy, 7-19 lesions, and 0.05-23cc, respectively.
Results: KBP planning achieved similar results for PTV CI100% and V100% [0.6% and 0.8% improvement, respectively] while reducing normal brain dose (V3Gy of 420cc and V10Gy of 31.9cc compared to 456cc and 35cc, respectively). Compared to MME, KBP reduced D0.03cc for brainstem (579cGy to 575cGy), optic nerves (191cGy to 144 cGy), chiasm (233cGy to 139cGy), eyes (137cGy to 84cGy), and lens (62cGy to 51cGy). KBP plans exhibited smaller variations in PTV coverage and OAR dose-volume criteria, suggesting greater reproducibility and consistency. KBP reduced the number of arcs (22.7 to 7.5), MU (17710 to 10877), and isocenters (2.6 to 1.67), thus reducing treatment delivery duration.
Conclusion: KBP can aid planners in delivering a highly conformal SRS plan for multiple brain metastases, while sparing normal brain and nearby OARs better, on average, than MME. Generating a robust, in-house SRS KBP model provides a viable option in planning cases involving multiple brain metastases.
Stereotactic Radiosurgery, Inverse Planning, Intensity Modulation