Room: Exhibit Hall
Purpose: GammaPod TPS uses pre-calculated dose kernels based on fixed-size breast cups and uniform density at 5mm control point resolution. The planning accuracy is impacted, if patient breast does not match the cups. We propose a novel dose kernel converter (DKC) as a fast and accurate dose engine for online GammaPod planning.
Methods: GammaPod delivery is a sequence of time modulated control points (CPs). At each CP, GammaPod generates a dose distribution (kernel) by radiation crossfire. The proposed DKC has the following steps: 1) for each collimator size (25mm and 15mm), we calculate a single reference dose kernel with CP at the target center using a fast graphic processing unit-based Monte Carlo (MC) dose calculation (~1 minute); 2) for a CP in the target, we calculate the mean radiological distance from the CP to all source positions using collapsed-cone ray-tracing; 3) for an arbitrary CP position, we generate its dose kernel by converting the reference kernel via shifting and scaling. Specifically, the reference kernel is shifted to the given CP and scaled by the ratio of mean tissue phantom ratio over all beam directions at the given CP to that at the reference CP. The mean tissue phantom ratio is calculated via mean radiological distance as calculated in Step 2. DKC generates dose kernels on the fly during plan optimization. We verified the accuracy of DKC using full MC dose calculation.
Results: For all studied patient CTs and tumor shapes, the agreement between DKC dose and MC dose was above 98% using the gamma passing criteria 1%/1mm, as the CPâ€™s in the target were in close proximity to the reference CP.
Conclusion: The proposed DKC directly uses patient CT to generate dose kernels at any CP efficiently without limitation of CP resolution, allowing accurate and fast online plan optimization.