(Wednesday, 7/17/2019) 8:30 AM - 9:30 AM
Room: Stars at Night Ballroom 2-3
Purpose: To rapidly simulate and analyze the probabilistic effect of various uncertainty sources in the radiation therapy on dosimetric and biological indices using high performance computation techniques.
Methods: The Radiation therapy robustness analyzer (RTRA) simulates daily patient setup error, independent rigid region-of-interest (ROI) motion, and patient specific delineation uncertainties. RTRA calculates probabilistic biological indices (e.g. E[TCP], E[NTCP]), as well as pDVH, pDMH, and dose-volume coverage maps for selected ROIs. RTRA is written in C++ and CUDA programing languages, and can be used in interactive (via GUI) and scripted modes. The required data sets are CT, structure set, and RT dose, which are supported in DICOM and Pinnacle formats. RTRA employs Monte-Carlo simulation techniques to simulate thousands of virtual treatment courses (VTC). In each VTC simulation, a unique random set of systematic errors are sampled, and in each treatment fraction ROI point clouds are transformed by a map both sampled from geometric uncertainty models. To minimize the memory usage, RTRA leverages highly accelerated incremental quantile determination techniques to estimate dose statistics at voxel-level which enables 3-D dose distribution at different confidence levels. In scripted mode, a detailed report is automatically generated for input data sets which enables running analysis in batch for a large population of patients.
Results: RTRA has been used in treatment plan robustness analysis of more than 500 different treatment plans in various treatment sites in our institution. A typical runtime to perform the 1000 virtual treatment simulation for 30 fractions with 10 ROIs is about 20 seconds with a single TITAN Xp, NVIDIA, GPUs.
Conclusion: RTRA assists clinicians to quickly analyze the robustness of treatment plans with respect to the relevant geometric uncertainties in external beam radiation therapy, and ultimately helps to determine the impact of individual uncertainty sources to the desired plan quality metrics.
Keywords
Not Applicable / None Entered.
Taxonomy
Not Applicable / None Entered.
Contact Email
