一种测量MR-linac配准误差的质量保证流程研究

A quality assurance process for measuring MR-linac registration deviation

目的 设计一种利用电子射野影像装置(EPID)实现精确摆位的方法,应用该方法探究MR-linac的配准误差并提出配准质量保证工作流程。方法 通过观察EPID图像中钨球位置的方法实现精确摆位,并选择Head模体验证该方法的精度。使用Zeus运动胸廓模体在5种不同运动模式下扫描模拟CT并分别生成参考计划。每组参考计划在模体位置和运动模式不变的情况下扫描治疗前MRI,并分别以脊髓和大体肿瘤体积(GTV)为感兴趣区域(ROI)与模拟CT配准并记录图像对齐的移动距离。在模体位置不变、运动模式改变的情况下扫描治疗前MRI进行配准以考察运动状态对配准误差的影响。对齐距离即配准误差(ΔX、ΔY和ΔZ),误差的正负值代表不同的方向,统计配准误差的大小范围并分析运动状态和ROI选择对误差的影响。结果 Head模体的配准误差为ΔX=1.10 mm,ΔY=1.80 mm,ΔZ=0.30 mm。Zeus模体的配准误差(ΔX,ΔY,ΔZ)最小值分别为-1.00、-1.40和-1.90 mm;最大值分别为0.60、6.30和-1.40 mm。ΔY的标准差最大,为2.36 mm;ΔZ的标准差最小,为0.19 mm。以GTV作为ROI时,所有方向误差的标准差均大于以脊髓作为ROI时的配准结果。MR与CT运动状态相同时的配准误差小于运动不同时的误差。结论 基于EPID图像辅助摆位的方法准确可靠,并提出了一种利用运动模体进行MR-linac配准质量控制的工作流程。MR-linac在上下方向的配准误差最不稳定,但总体误差范围是可接受的。

Objective To design a method to achieve accurate setting-up using electronic field imaging device(EPID),explores the registration error of MR-linac by using this method, and proposes the registration quality assurance(QA) workflow.Methods Accurate setting-up was achieved by observing the position of tungsten balls in EPID image, and the Head phantom was selected to demonstrate the accuracy of the method.The simulated CT scans of Zeus motion thoracic phantom in 5 different motion modes were scanned and the reference plans were generated respectively.Pre-treatment online MRI was scanned for each reference plan with the phantom position and motion modes unchanged.Moreover, the spinal-cord and gross tumor volume(GTV) were respectively used as regions of interest(ROI) to register with the simulated CT and the shift distance of image alignment was recorded.The pre-treatment MRI was also scanned for registration under the condition that the phantom position was unchanged and the motion mode was changed to investigate the influence of motion modes on registration error.The alignment distance is the registration error(ΔX,ΔY,ΔZ).The positive and negative values of the error represent different directions. The range of registration errors was counted.The influence of motion mode and ROI selection on error were analyzed.Results The registration errors of the Head phantom wereΔX=1.10mm,ΔY=1.80 mm,andΔZ=0.30 mm.The minimum registration errors (ΔX,ΔY,ΔZ)of Zeus phantom were-1.00,-1.40,-1.90mm and the maximum values were 0.60,6.30,-1.40mm,respectively.The standard deviation ofΔYwas the largest 2.36mm,and the average standard deviation ofΔZ was the smallest 0.19mm.When GTV was used as ROI,the standard deviation of errors in all directions was greater than the registration results when spinal-cord was used as ROI.The registration error of MR and CT with the same motion mode was smaller than that with different motion modes.Conclusions The setting-up method based on EPID image assistance is accurate and reliable,and a workflow of MR-linac registration quality control using moving phantom is proposed.MR-linac has the most unstable registration error in the inferior and superior direction,but the overall error range is acceptable.