基于Web的剂量体积直方图数据自动提取工具开发及验证

Development and validation of Web-based automated dose-volume histogram data analyzer

目的:开发并测试一套基于Web的剂量体积直方图(DVH)数据自动提取工具。方法:(1)采用Django应用框架和Python编程语言,设计一套基于Web的DVH数据自动提取工具。(2)利用自动工具分析从Eclipse计划系统中导出的30例相同类型计划的DVH表单数据,并采用人工方法读取DVH图中的相应参数作为对比,分析其耗时、准确性等方面的表现及误差产生原因。结果:自动提取DVH数据的效率远高于人工分析,且正确率更有保障。对于计划靶区体积的均匀性指数、股骨头和膀胱D_(50%)和平均剂量等参数,自动与人工提取的差异极小(误差≤0.01%,P>0.05)。但对于适形指数(CI)值的计算,由于计划系统空间采样算法的原因使得基于等剂量线结构转换测量以及DVH表单数据分析之间的结果存在较大差异[CI_PGTV平均相差(2.60±1.04)%,CI_PTV平均相差(0.66±0.29)%,P<0.001],但Web工具采用的DVH分析结果更加接近Eclipse自动生成的CI值,且有效避免了后者一次只能计算一个CI值的缺陷。结论:本工作开发的基于Web的工具可以对海量DVH数据进行高效、准确的自动统计,且具有跨平台应用等优势。

Objective To develop and validate a Web-based automated dose-volume histogram（DVH） data analyzer. Methods Based on the Django framework and Python programming language, a web-based automated DVH data analyzer was designed.The DVH in tabular format of 30 similar plans were exported from Eclipse planning system by using automated analyzer; the corresponding parameters of DVH were extracted manually. And the exported data were compared with the extracted data to analyze the efficiency, accuracy and the reasons of errors. Results Compared with manual method, the automated DVH data analyzer showed superior efficiency and accuracy. For the parameters such as HI_PTV, D_（50%）and mean dose for the femoral head and urinary bladder, the disparities between manual and automatic extraction were negligible（Error≤0.01%, P0.05）. However,significant differences in the calculation of conformity index（CI） were observed between the isodose conversion method and the DVH extraction, which was ascribed to the spatial sampling algorithms of planning system. The mean difference in CI_PGTV and CI_PTVwere respectively（2.60±1.04）% and（0.66±0.29）%, P0.001. The results of DVH extraction adopted by the Web-based analyzer were more close to the CI values auto-generated by Eclipse. And the Web-based analyzer effectively conquered the limitation of Eclipse which only calculated a CI value at a time. Conclusion The developed Web-based DVH analyzer can autoanalyze large amount of DVH data efficiently and accurately, with the advantage of cross-platform application.