Program Information
Ultra-Fast SRS Planning for Trigeminal Neuralgia Using Quadratic Optimization
R Ma1*, S Zhou1 , B Lu2, D Zheng1 , G Yan2 , Y Lei1 , Q Fan1 , (1) University of Nebraska Medical Center, Omaha, NE, (2) University of Florida, Gainesville, FL
Presentations
WE-RAM3-GePD-T-6 (Wednesday, August 2, 2017) 10:30 AM - 11:00 AM Room: Therapy ePoster Lounge
Purpose: SRS is a popular treatment option for trigeminal neuralgia (TN) due to its noninvasiveness. Simulation, planning, QA, and delivery are usually accomplished in a single day. Out of this fast-paced workflow, planning remains most challenging as seeking an optimal dose distribution in a short timeframe requires a substantial level of experience. In this work, we tackle this challenge by replacing the empirically iterative process of finding appropriate beam weighting with deterministic optimization.
Methods: SRS for TN is usually planned with seven to eight circular arcs. The determination of relative beam weighting of each arc is formulated into a convex optimization problem. Specifically, we search for a solution that minimizes the L2-norm of the difference between the calculated dose (denoted as Ax) and the prescribed dose (denoted as d) for given constraints (i.e., x>0). A is a matrix whose column corresponds to the dose kernel contributed by individual arc. x is a vector representing the optimal beam weighting to be determined. A regularization term is utilized to ensure the weighting distribution is relatively smooth. The optimization problem is solved using the MOSEK optimizer (MOSEK ApS, Denmark). For evaluation, 10 LINAC-based SRS cases planned with BrainLab iPlan were enrolled for a retrospective study. We re-calculated the clinically approved plan with the optimized beam weighting for comparison.
Results: With our proposed method, all 10 cases were able to achieve the prescription goal similarly as the clinical plans. The maximum dose to brainstem has been reduced by an average of 4.6% of prescription dose. The overall treatment planning time has been shortened from the order of hours to minutes.
Conclusion: By formulating the SRS planning process into a convex optimization problem, the planning time could be substantially reduced while achieving a similar or better dose distribution.
Contact Email: