Impact of inconsistent dose calculation algorithms on modeling and use of automated treatment planning for lung stereotactic body radiotherapy

曲宝林,戴相昆,解传滨,刘宏嘉,王海洋,吴昊
DOI: https://doi.org/10.3760/cma.j.cn115668-20200403-00079
2021-01-01
Abstract:Objective:To investigate the impact of different dose volume calculators on RapidPlan modeling and use of automated treatment planning for lung stereotactic body radiotherapy (SBRT).Methods:Twenty-seven historical plans calculated with Anisotropic Analytical Algorithm (AAA) were used as the training set of model AAA. These 27 plans were recalculated using Acuros XB (AXB) algorithm as the training set of model AXB. Another 20 plans which were with different prescribed doses and independent from the modeling were selected to validate the models. Dosimetric comparisons were performed for the training and validation sets respectively between the two models.Results:(1) Comparison of training sets between the original and recalculated models showed that the AAA had higher target dose coverage on mean dose volume histograms (DVHs) , and the dose differences at organs at risk (OARs) were not significant between models. Except for V 5 Gy and V 10 Gy (relative volumes receiving at least 5 Gy and 10 Gy respectively) of lung and the maximum dose (D max) of trachea, the differences of other dosimetric parameters suggested that the planned doses calculated by AXB algorithm was generally lower than those by AAA algorithm. In particular, the difference in mean conformity index (CI) of the planning target volume (PTV) reached 13.98%, and that in PTV_D 95% (dose received by 95% of PTV) was up to 3.50%. (2) Comparison of validation sets showed that, except for the higher dosimetrics with AXB optimization in high-dose area of the target, there was no significant difference in dosimetry at other areas. The majority of OARs did not differ significantly between AXB and AAA models, with the maximum difference being no more than 5.64%. Conclusion:It is not necessary to recalculate the training set planned with a different dose volume calculator during RapidPlan modeling. On the other hand, to preserve the geometric diversity and broad representation of the model, it is not recommended to exclude historical plans calculated by different algorithms from the training set.
What problem does this paper attempt to address?