In this paper, a GPU accelerated robust optimizer for proton intensity modulated radiation therapy (IMRT) is developed to reduce the impact of range uncertainty and target localization bias. The objective function used in the robust optimization model includes 9 kinds of boundary dose objectives, which are: no deviation, 2 kinds of range deviation (long and short), 6 kinds of setup uncertainty (2 kinds of positive and negative deviation in front and back, side, up and down incident directions). Firstly, the dose contribution matrix of target area and organs at risk is calculated by pencil beam algorithm, and then the conjugate gradient method is used to optimize the objective function to meet the constraint conditions. Three clinical cases of head and neck, lung and prostate were used to test the performance of the optimizer. Compared with the traditional PTV based proton intensity modulated radiotherapy (IMRT) planning, the robust optimizer can optimize the treatment plan which is more insensitive to the range uncertainty and setup error, so that the target area can achieve high dose uniformity and the organs at risk (OARs) can be better protected. After 100 iterations, the optimization time of the three cases is about 10 seconds. The results show that the GPU accelerated robust optimizer can design a high robust proton therapy plan in a short time, so as to improve the reliability of proton therapy.
The uncertainty model is based on the standard optimization function, which includes the range uncertainty and target location error. Conjugate gradient (CG) method is used to optimize the uncertainty model, and GPU is used to accelerate the matrix vector multiplication.
Robust optimization model processing
In this paper, three cases are used to test the effectiveness of the robust optimizer, and the data is good
| # | number | name | type |
| 1 | 2017YFC0107500 | Research and implementation of a new conformal intensity modulation technique for multi-particle biology guided by multi-modes | National key R & D plan |
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
CSTR:11738.11.NCDC.IMP.DB4089.2023
10.12072/ncdc.imp.db4089.2023
