Room: Exhibit Hall
Purpose: We have been developing a method to predict gamma passing rate (GPR) by using the dose uncertainty (DU) accumulation model. This method uses the DU distribution of the evaluated data together with learning data (LD) of gamma and DU distributions of head-and-neck (H&N) data. This method was tested for H&N data by the leave-one-out cross-validation (LOOCV) and showed good performance. The purpose of this study is to investigate the application of LD for other treatment sites. In this study, GPR of prostate IMRT was examined.
Methods: IMRT treatment plans of 10 H&N and 8 prostate cases were created in XiO treatment planning system. All plans used 9 SMLC beams of ONCOR. The planar dose distribution of each beam was measured using MapCHECK. LD of gamma and DU distributions were created using the H&N data. GPR of each beam in H&N and prostate IMRT was predicted using its own DU distribution together with LD. For H&N beams, LOOCV was used. Criteria of the predicted GPR (pGPR) corresponding to the measured GPR (mGPR) > 90% was calculated with the confidence level of 99.9%.
Results: In both treatment sites, a good proportionality between pGPR and mGPR was obtained. The calculated standard deviation (SD) of pGPR-mGPR was 2.2% and 1.9% for H&N and prostate, respectively, for 3mm-3% tolerance. The lower boundary of pGPR corresponding to mGPR>90% (-3.1SD corresponding to 99.9% confidence interval) was 97.1% and 91.1% for H&N and prostate, respectively. Among the beams examined in this study, 74% and 64% of H&N and prostate cases met these criteria. Therefore, it is expected that this pGPR may help to omit measurements.
Conclusion: The pGPR developed in this study showed good accuracy for prostate IMRT. Future prediction accuracy is expected to be improved further by considering correspondence to three dimensions in the future.
Penumbra, Quality Assurance, Radiation Therapy