Room: Exhibit Hall | Forum 7
Purpose: GammaKnife inverse planning involves determining the position, cone, sector, and intensity of each shot for an unknown number of shots, resulting in a non-convex mixed-integer problem, and becoming intractable when the required number of shots is large, e.g. with a large volume or number of tumors, such as multiple brain metastases. We proposed an efficient and effective linear programming (LP)-based approach to optimize all these parameters.
Methods: We proposed a two-step solution. First, we determined the number and positions of the shots, using LP with target voxels as candidate positions and a coarse dose grid. The optimized positions were sparse, and the number of shots was further reduced by removing those with low intensity. Then, we performed full optimization using LP at these shot positions to obtain intensity for 24 combinations of cones and sectors by minimizing the total beam-on time, dose to normal tissues, or a combination of both, subject to prescription dose constraints to the targets. We applied this inverse planning approach to GammaKnife for multiple brain metastases and compared plan quality and delivery time of the results and clinical plans.
Results: The first step was able to find optimal, sparse solutions within a couple minutes. Depending on the optimization objectives, the optimized beam-on times may be shorter or dose to normal tissues may be lower compared to the clinical plans. For the studied case, the total beam-on time was comparable, but the conformity was improved by 6%, compared to the clinical plans.
Conclusion: The two-step approach alternates dimension reduction in dose grid and isocenter candidates, making the problem tractable, and LP was able to find sparse solutions. The proposed inverse planning can efficiently determine the treatment isocenter, cone/sector type, and intensity for each shot, and thus provide quality plans with minimum beam-on time.
Funding Support, Disclosures, and Conflict of Interest: This work is partially supported by NIH R01 CA235723 and the seed grant of Radiation Oncology Department at University of Texas Southwestern Medical Center.