Li S, Lu M, Kim J, Glide-Hurst C, Chetty I, Zhong H
Henry Ford Health System, Detroit, MI.
Med Phys. 2012 Jun;39(6Part21):3875. doi: 10.1118/1.4735823.
Purpose Clinical implementation of adaptive treatment planning is limited by the lack of quantitative tools to assess deformable image registration errors (R-ERR). The purpose of this study was to develop a method, using finite element modeling (FEM), to estimate registration errors based on mechanical changes resulting from them. Methods An experimental platform to quantify the correlation between registration errors and their mechanical consequences was developed as follows: diaphragm deformation was simulated on the CT images in patients with lung cancer using a finite element method (FEM). The simulated displacement vector fields (F-DVF) were used to warp each CT image to generate a FEM image. B-Spline based (Elastix) registrations were performed from reference to FEM images to generate a registration DVF (R-DVF). The F- DVF was subtracted from R-DVF. The magnitude of the difference vector was defined as the registration error, which is a consequence of mechanically unbalanced energy (UE), computed using 'in-house-developed' FEM software. A nonlinear regression model was used based on imaging voxel data and the analysis considered clustered voxel data within images. Results A regression model analysis showed that UE was significantly correlated with registration error, DVF and the product of registration error and DVF respectively with R̂2=0.73 (R=0.854). The association was verified independently using 40 tracked landmarks. A linear function between the means of UE values and R- DVF*R-ERR has been established. The mean registration error (N=8) was 0.9 mm. 85.4% of voxels fit this model within one standard deviation. Conclusions An encouraging relationship between UE and registration error has been found. These experimental results suggest the feasibility of UE as a valuable tool for evaluating registration errors, thus supporting 4D and adaptive radiotherapy. The research was supported by NIH/NCI R01CA140341.
自适应治疗计划的临床应用因缺乏评估可变形图像配准误差（R-ERR）的定量工具而受到限制。本研究的目的是开发一种基于有限元建模（FEM）的方法，根据配准误差导致的力学变化来估计配准误差。方法：开发了一个用于量化配准误差与其力学后果之间相关性的实验平台，具体如下：使用有限元方法（FEM）在肺癌患者的CT图像上模拟膈肌变形。模拟的位移矢量场（F-DVF）用于对每个CT图像进行变形，以生成有限元模型图像。从参考图像到有限元模型图像进行基于B样条（Elastix）的配准，以生成配准DVF（R-DVF）。从R-DVF中减去F-DVF。差异矢量的大小被定义为配准误差，这是使用“自行开发的”有限元软件计算的机械不平衡能量（UE）的结果。基于成像体素数据使用非线性回归模型，并且分析考虑了图像内的聚类体素数据。结果：回归模型分析表明，UE分别与配准误差、DVF以及配准误差与DVF的乘积显著相关，R̂2 = 0.73（R = 0.854）。使用40个跟踪的地标独立验证了这种关联。已经建立了UE值的平均值与R-DVF*R-ERR之间的线性函数。平均配准误差（N = 8）为0.9毫米。85.4%的体素在一个标准差内符合该模型。结论：发现UE与配准误差之间存在令人鼓舞的关系。这些实验结果表明，UE作为评估配准误差的有价值工具具有可行性，从而支持四维和自适应放射治疗。本研究得到了美国国立卫生研究院/国立癌症研究所R01CA140341的支持。
