Quality improvement of automatic planning for nasopharyngeal carcinoma by iteratively updating model(PDF)
《中国医学物理学杂志》[ISSN:1005-202X/CN:44-1351/R]
- Issue:
- 2022年第2期
- Page:
- 139-145
- Research Field:
- 医学放射物理
- Publishing date:
Info
- Title:
- Quality improvement of automatic planning for nasopharyngeal carcinoma by iteratively updating model
- Author(s):
- WANG Yixuan1; 2; ZHU Jinhan2; ZHANG Dandan2; DENG Xiaowu2; ZHOU Linghong1
- 1. School of Biomedical Engineering, Southern Medical University, Guangzhou 510515, China 2. Department of Radiation Oncology, Sun Yat-sen University Cancer Center/State Key Laboratory of Oncology in South China/Collaborative Innovation Center for Cancer Medicine, Guangzhou 510060, China
- Keywords:
- Keywords: nasopharyngeal carcinaoma automatic planning model iteration
- PACS:
- R318；R811.1
- DOI:
- DOI:10.3969/j.issn.1005-202X.2022.02.002
- Abstract:
- Abstract: Objective To establish a new iteratively updating method for improving the quality of Varian eclipse Rapid Plan automatic planning model in radiotherapy planning for nasopharyngeal carcinoma (NPC), and to verify its reliability in clinical application. Methods The radiotherapy plans were randomly selected from 117 NPC patients, and the plan quality indexes were determined and scored using evaluation method and RTOG protocol 0615 report.?he plans which were used for model training were screened by score threshold and each iteratively updating model was used to assist the redesign of the low-score plans which were screened out. The score threshold was gradually improved by iterations, and the final model was obtained after 3 iterations.?nother 20 NPC test cases were selected to verify the changes of the model after iteration, and the comparison between the final model and the original manual plan was carried out. Results The models A1, A2 and A3 were obtained by 3 iterations.?ith the iteration, the scores showed an upward trend, and the score of the final model A3 was significantly higher than the score of the original manual plan (P<0.05).?he target dose coverage of the 3 models all met the clinical requirements, and their differences in the mean value of V70 Gy of PTVnx, V60 Gy of PTV1 and V54 Gy of PTV2 were not more than 1%, but the Dmax to the spinal cord and the Dmax to optic chiasma had an obvious downtrend with the iterations of the model.?ompared with those in the original manual plan, the V54 Gy of brainstem, the Dmax to spinal cord, the Dmean to left parotid gland, the V60 Gy of left and right temporal lobe, and the Dmax to optic nerve and optic chiasma in the final model A3 were decreased significantly, with statistical differences (P<0.05).?onclusion The automatic planning for NPC based on iteratively updating model is explored in the study, and it is found that the model quality and prediction ability can be effectively improved through continuously improving the training set, thereby effectively enhancing the quality of automatic planning and assisting physicists or dosimetrists in designing high-quality plans.
Last Update: 2022-03-07