摘要
目的对使用AccuContour软件自动勾画全脑全脊髓放射治疗(CSI)患者靶区和危及器官(OARs)进行风险评估和风险应对。方法选取2012年1月至2022年12月于医院接受全脑全脊髓放射治疗(CSI)的40例患者为研究对象,其中仰卧位和俯卧位各20例,且同一体位组内成人和儿童各10例。手动和自动勾画4组患者靶区和OARs。界定风险准则,勾画结果以德尔菲法进行后果分析,对靶区和OARs做出风险分型,并逐例逐层回顾统计,通过危险与可操作性分析法进行可能性分析,对勾画差异做出归因分类。结果脑、肝脏、左右肺、左右肾、心脏和左右眼球为鲁棒型;甲状腺、左右腮腺为平衡型;喉、气管、食管、胃、脊髓腔、左右视神经和左右晶体为审慎型。自动勾画和手动勾画的差异归因可分为Ⅰ~Ⅶ类。根据器官分型和归因分类,对开发方和应用方提出不同的风险应对建议。结论AccuContour软件应用于CSI患者的靶区和OARs自动勾画具有较高的临床价值,但需完善前期风险评估和策略应对,确保智能勾画的安全、可靠、可控及可持续发展。
Objective Perform risk assessment and treatment for the use of AccuContour software to automatically delineate the craniospinal target volume and organs at risk(OARs).Methods 40 patients with craniospinal irradiation(CSI)in our hospital were selected,including 20 supine and 20 prone patients,and 10 children and 10 adults in each position group.The CT images and existing structure files(RS)of all patients in the four groups were imported into Eclipse,and the target volume and OARs were manually delineated and modified according to the guidelines.Transfer CT and the manually contoured RS files into AccuContour software,which automatically delineated using the CSI template provided by the software.For the delineated results,firstly,the risk criteria are defined through consequence analysis by DELPHI method,and the risk classification is made for the target volume and OARs by using the Daisy Similarity Coefficient(DSC)and 95th percentile Hausdorff distance(HD95)evaluation parameters.Then,case-by-case and layer-by-layer review statistics are adopted in each OARs to summarize and classify the differences in the results through likelihood analysis by Hazard and Operability Analysis(HAZOP)method.Further develop clinical action strategies and software upgrade response accordingly.Results Brain,liver,left and right lungs,left and right kidneys,heart,and left and right eyeballs are Robust-type,accept or review.Thyroid and left and right parotid glands are of Balanced-type,intervene and adjust.Larynx,trachea,esophagus,stomach,spinal cord cavity,left and right optic nerves,and left and right lenses are Cautious-type,control and modify.The possible reasons for differences between automatic and manual delineation could be classified into seven types,and the upgrade response for typesⅠ-Ⅲwas Optimization-level and typesⅣ-Ⅵwas a Correction-level.Conclusion Using AccuContour software to automatically delineate target areas and organs at risk in CSI patients has great value in clinical applications,but it requires early risk assessment and risk treatment to ensure safe,reliable,controllable,and sustainable development of intelligent delineation.
作者
李江山
杨志宇
蔡鸿斌
林智星
黄妙云
Li Jiangshan;Yang Zhiyu;Cai Hongbin;Lin Zhixing;Huang Miaoyun(Fujian Medical University Union Hospital,Fujian Key Laboratory of Intelligent Imaging and Precision Radiotherapy for Tumors(Fujian Medical University),Clinical Research Center for Radiology and Radiotherapy of Fujian Province(Digestive,Hematological and Breast Malignancies),Fuzhou Fujian 350001,China)
出处
《医疗装备》
2023年第18期1-7,共7页
Medical Equipment
基金
福建省卫生健康中青年骨干人才培养项目(2019-ZQN-36)
福建省科技项目(2019Y0020)。