一种大分割立体定向放射治疗计划质量评估方法  

A Quality Evaluation Method for Hypofractionated Stereotactic Radiotherapy Plan

作　　者：殷皓泽 张婷 李克[1] 周元波[1] 李丹[1] 汪森 邱凌平[1] Yin Haoze;Zhang Ting;Li Ke;Zhou Yuanbo;Li Dan;Wang Sen;Qiu Lingping(The First Affiliated Hospital of Nanchang University,Nanchang Jiangxi 330006,China)

机构地区：[1]南昌大学第一附属医院,江西南昌330006

出　　处：《医疗装备》2023年第22期6-10,共5页Medical Equipment

基　　金：江西省教育厅科学技术重点研究项目(GJJ2200118)。

摘　　要：目的提出一种针对大分割立体定向放射治疗(HSRT)计划的质量评估方法,通过对评估参数进行建模拟合,为调整治疗计划指导方向。方法选取2021年1—12月医院收治的20例病灶直径>2 cm的脑转移瘤患者,基于Monaco5.11治疗计划系统,从肿瘤控制率和正常组织并发率出发,以适形度指数[CI,包括肿瘤放射治疗组(RTOG)CI和Paddick CI]、梯度指数(GI)、质量指数(QI)、均匀性指数(HI)、V_(24.4)、V_(12)、V_(18)、V_(23)及计划靶区大小(PTVbin)等为质量评估参数,利用集成学习中的Bagging算法拟合参数形成一系列基学习器,整合对应权重建立质量评估集群分类学习模型。结果所选评估参数对HSRT计划质量的影响程度排序依次为V_(12)、V_(24.4)、PTV_(bin)、GI、CI_(Paddick)、V_(23)、QI、V_(18)、CI_(RTOG)、HI。其中,V_(12)对计划质量的影响最大,占16.2%;HI影响最小,占3.8%。结论该方法以多因子模型为核心,利用集成学习将计划质量评估过程模拟为多个基学习器,组合预测结果确认各个参数对计划质量的影响程度,明确了计划设计优化的目标,为制定高质量治疗计划提供了统计学证据。Objective With the proposal of a quality evaluation method for hypofractionated stereotactic radiotherapy(HSRT)plans,the evaluation parameters were modeled and fitted to guide the adjustment of treatment plans.Methods With the selection of patients with brain metastases with the lesion diameter greater than 2 cm admitted to the hospital from January to December 2021,based on the Monaco5.11 treatment planning system,from the perspective of tumor control rate and normal tissue complication rate,the fitness index[CI,included radiation therapy oncology group(RTOG)CI and Paddick CI],gradient index(GI),quality index(QI),homogeneity index(HI),V_(24.4),V_(12),V_(18),V_(23),and bins of planning target volume(PTVbin),were used as quality evaluation parameters.In addition,the Bagging algorithm in ensemble learning was utilized to form a series of base learners using fitting parameters,and corresponding weights were integrated to establish a quality evaluation cluster classification learning model.Results The ranking of the impact of the selected evaluation parameters on the quality of the HSRT plan was V_(12),V_(24.4),PT_(Vbin),GI,CI_(Paddick),V_(23),QI,V_(18),CI_(RTOG) and HI.Among them,V_(12) had the greatest impact on plan quality,accounted for 16.2%;HI had the smallest impact,accounted for 3.8%.Conclusion With the multi-factor model as the core,ensemble learning is utilized to simulate the process of plan quality assessment into multiple base learners,combining prediction results to confirm the degree of influence of various parameters on plan quality,clarifying the optimization objectives of plan design,and providing statistical evidence for the development of high-quality treatment plans.

关 键 词：大分割立体定向放射治疗 质量评估 集成学习 基学习器 脑转移 肿瘤控制率 正常组织并发率 

分 类 号：R730.55[医药卫生—肿瘤]