Room: ePoster Forums
Purpose: We aim to develop a novel method to generate an optimized learning data (LD) for predicting gamma passing rate (GPR) by using dose uncertainty potential (DUP) accumulation developed in our recent study [Shiba et al, Med Phys 46, 999-1005 (2019)], in which a hypothesis that the DUP is a good metric for the dose accuracy.
Methods: 38 head-and-neck IMRT treatment plans were created in XiO treatment planning system. All plans were created using 9 SMLC beams of ONCOR. Verification plans were created and measured by Delta4. The 3D DUP manifesting on the field edge was generated by assuming the Gaussian distribution in the lateral direction and the longitudinal profile with the same shape as the depth dose. The LD for predicting GPR (3%/ 3 mm, 3%/ 2 mm, and 2%/ 2 mm) was assumed by an exponential function: aGPR(DUP) = exp[-q DUP]. The coefficient q was optimized so that the difference between the predicted GPR (pGPR) and the measured GPR (mGPR) was minimized (dGPR = pGPR - mGPR). Standard deviation (SD) of dGPR was evaluated for the optimized LD.
Results: It was confirmed that the coefficient q was larger for tighter tolerance. This result corresponds to the expectation that the larger attenuation of the aGPR(DUP) for tighter tolerance. A good proportionality between pGPR and mGPR was obtained. The SD of dGPR was 2.1, 3.8, 6.0% for the tolerance of 3%/3 mm, 3%/2 mm, 2%/2 mm.
Conclusion: A technique to optimize the LD was developed for DUP-based prediction of the 3D GPR of head and neck IMRT. In the future plan, we will conduct the cross-validation test and evaluate the performance of our 3D GPR prediction.