A feasibility and dosimetric study of configuring a Varian RapidPlan model based on Elekta cases

李正贤,闫相东,王博生,岳海振,李晨光,赵亮,黄宇亮,吴昊,张艺宝

DOI: https://doi.org/10.3760/cma.j.issn.1674-1927.2020.01.002

2020-01-01

Abstract:Objective:To investigate the potential of multi-platform modeling using pre-surgical rectal cancer SIB-VMAT plans as example, and to evaluate the dosimetric performance of the new model with reported Beijing Cancer Hospital (BJH) model as the reference.Methods:The images, structures and dose matrix of 68 cases treated on Elekta Linac were transferred from Monaco TPS in Jilin Guowen Hospital (GWH) and imported into RapidPlan database, replacing the original 81 training cases of BJH model without changing other parameters such as rules of objective generation. A GWH model was obtained through statistical verification and outlier management using Model Analytics after refitting. Another 11 validation cases from GWH were selected to compare the optimization performance of the two models based on RTOG0822 protocols in terms of V 35 Gy, V 40 Gy and V 45 Gy for small intestines, V 40 Gy and V 45 Gy for femoral heads, V 40 Gy and V 45 Gy for urinary bladder. Results:RapidPlan model was successfully constructed using Elekta data and the re-optimized plans for all the validation cases passed RTOG0822 criterion in a single run, with significant improvement compared with the 68-case training set representing the average quality of the manual plans. BJH model better spared urinary bladder (9.33% lower from 23.66 to 21.64 for V 35 Gy; 5.97% lower from 14.20 to 13.40 for V 45 Gy; all P<0.05) , small bowel (20.04% lower from 13.48 to 11.23 for V 35 Gy, 20.53% lower from 8.69 to 7.21 for V 40 Gy, and 26.84% lower from 4.82 to 3.80 for V 45 Gy; all P<0.05) , and femoral heads (71.21% lower from 1.13 to 0.66 for V 40 Gy, P<0.05, and 100% lower from 0.10 to 0.05 for V 45 Gy, P=0.08) . All differences were statistically significant ( P<0.05) except for V 45 Gy of femoral heads. Conclusions:Rapidplan model can be constructed and applied cross platforms with good tolerance to different prescriptions and accelerators.