Automatic planning of IMRT for rectum cancer based on optimization parameters tree search algorithm
10.3760/cma.j.issn.0254-5098.2021.01.014
- VernacularTitle:基于优化参数树搜索算法的直肠癌调强放疗自动计划
- Author:
Hanlin WANG
;
Jiacheng LIU
;
Kaining YAO
;
Ruoxi WANG
;
Jian ZHANG
;
Haizhen YUE
;
Yibao ZHANG
;
Hao WU
- From:
Chinese Journal of Radiological Medicine and Protection
2021;41(1):66-73
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To solve the problems in intensity-modulated radiation therapy (IMRT) planning, such as large labor cost and high dependence on the experience of physicists and great inconsistency in the quality of plan, and to discuss an unsupervised automatic treatment planning procedure of IMRT.Methods:The eclipse scripting application programming interface (ESAPI) within the Eclipse treatment planning system (TPS) 15.6 and optimization parameters tree search algorithm (OPTSA) were used to emulate and realize the whole planning process. Interacted with the TPS through ESAPI, relevant dosimetric parameters were input and output. The OPTSA evaluated the plan qualities based on dosimetric parameters of the targets and organs at risk (OARs) and iteratively adjusted the optimization objective parameters to achieve a progressively improving IMRT plan. In order to verify the effectiveness of the automatic planning, twenty historical rectum cancer cases were selected from the clinical database, and the dose distribution and specific dosimetric parameters were compared between the plans generated by the OPTSA and the manual plans under the same constraints.Results:All the auto plans have met clinical requirements. Furthermore, 90% and 10% of the auto plans were deemed as clinically improved and equally compared with the manual plans, respectively. The average CI for the PTV was 0.88 and 0.80 for the auto and manual plans respectively. Compared with the manual plans, the mean doses of all the OARs in the auto plans were reduced by 11% in average. The average elapsed time of automatic planning and manual planning was (28.15±3.61) and (36.7±4.6) min, respectively.Conclusions:The plans created by the proposed algorithm have been shown to be at least as good as the manual plans. In addition, this method can shorten the labor time in plan designing while ensuring the plan quality and consistency of the plan.
