编译工作流自动生成调强计划剂量限制结构的研究

Research on automatic dose limited structures generation by workflow tool of MIM software platform

目的探讨利用MIM V6.5软件平台的工作流工具,实现调强放射治疗计划剂量限制结构自动勾画的可行性。方法随机选择行放射治疗的不同部位肿瘤患者20例,其中男性8例,女性12例;年龄30~79岁,平均年龄55岁。鼻咽癌、食道癌、胃癌及宫颈癌各5例。根据调强放射治疗计划剂量优化的要求,确定所需勾画的剂量限制结构,利用MIM V6.5软件工具编译工作流,并分别采用工作流批处理方法(自动勾画)和TPS Eclipse V11.0提供的勾画工具(手动勾画)生成剂量限制结构,计算两种勾画方法生成的剂量限制结构体积重叠的Dice系数,并对比勾画时间的差异。结果将操作步骤按逻辑关系排序后,依据MIM V6.5软件提供的工作流工具编译工作流,分别生成ATUO CONTOUR NPC、ATUO CONTOUR ESOPHAGUS、ATUO CONTOUR STOMACH、ATUO CONTOUR CERVIX 4个工作流。自动勾画与手动勾画生成的所有剂量限制结构的Dice系数均大于0.9,能够满足临床计划制作的要求。自动勾画最长耗时为3.73 min,相比手动勾画在鼻咽癌、食道癌、胃癌、宫颈癌勾画时间平均分别节省约89%、50%、49%、63%,差异有显著统计学意义(P<0.01)。结论使用工作流可实现剂量限制结构勾画的自动化,提高工作效率,并为进一步研究治疗计划系统智能化提供有效借鉴。

Objective To evaluate the feasibility of automatic dose limited structure（Dis） generation in intensity-modulated radiotherapy（IMRT） plans with the workflow tool based on MIM software V6.5. Methods A total of 20 tumor patients in different body parts were randomly enrolled, which included 8 males and 12 females, aged 30 - 79 years old with mean age of 55 years old. Among these cases, there were 5 of nasopharyngeal cancer, 5 of esophageal cancer, 5 of gastric cancer and 5 of cervical cancer. For IMRT dose optimization, the DIS was derermined to compile workflow by MIM software, and then generated DIS by workflow batch method（automatic contour） and Eclipse treatment plan system（TPS, manual contour）（V11.0）, respectively. The Dice coefficient was calculated and time differences of the dose-limiting structures were compared generated between 2 methods. Results Sorted the steps in logical relationship, four workflows were generated by workflow tool of MIM software, which were named as ATUO CONTOUR NPC, ATUO CONTOUR ESOPHAGUS, ATUO CONTOUR STOMACH and ATUO CONTOUR CERVIX, respectively. The Dice coefficients of all dose limiting structures generated by automatic contour and manual contour was greater than 0.9, which met the requirements of clinical planning. The longest time of automatic contour was 3.73 minutes, compared with manual contour in nasopharyngeal carcinoma, esophageal cancer, gastric cancer and cervical cancer, which average saved about 89 %, 50 %, 49 % and 50 %, respectively, the difference was statistically significant （P 〈 0.01）. Conclusion It is demonstrated that MIM software in compile workflow could automatically generate DLS and effectively reduce planning time, and provide an effective reference for the further study on intelligent TPS system.