Federal Institute of Technology.
J Appl Clin Med Phys. 2013 Sep 6;14(5):222-30. doi: 10.1120/jacmp.v14i5.4374.
Volumes change during fractionated radiotherapy (RT). We investigate a tool based on the Hierarchical Enhanced Registration Algorithm (HERA) to project a 3D segmentation set of the prostate into the subsequent imaging sets at any time point during RT by using intensity-based image registration techniques. Sequential CT sets during RT at 15, 30, 45, and 60 Gy of two patients were used. Five expert clinicians outlined the prostate in a blinded fashion, defining intraobserver and interobserver variability on a set of 35 and 25 scans, respectively. The observer variability and positioning for manual correction was compared to both affine and elastic image registration-based contour propagation. The overall mean error of the registration-based correction of the planning target volume was comparable to the interobserver variability of manual target volume definition. The correction by affine image fusion was inferior to the results of elastic registration. The maximal deviation for the interobserver segmentation was 15.4 mm, 10.5 mm for the affine and 8.0 mm for the elastic registration. The mean interobserver variability was 1.5 (± 1.4) mm, 2.8 (± 2.3) mm for the affine, and 2.2 (± 1.9) mm for the elastic registration. Intensity-based elastic registration of deformable anatomical structures with HERA is suitable for the assessment of changes of prostate volumes for the purpose of target propagation and adaptive radiotherapy.
在分割放疗 (RT) 期间,体积会发生变化。我们研究了一种基于分层增强配准算法 (HERA) 的工具,该工具可使用基于强度的图像配准技术,将前列腺的三维分割集在 RT 期间的任何时间点投影到随后的成像集中。使用了两位患者在 15、30、45 和 60 Gy 时的连续 CT 集。五名专家临床医生以盲法方式勾勒出前列腺,分别在 35 次和 25 次扫描上定义了观察者内和观察者间的可变性。将手动校正的观察者可变性和定位与仿射和弹性图像配准的轮廓传播进行了比较。基于配准的校正的规划靶区总体平均误差与手动靶区定义的观察者间可变性相当。仿射图像融合的校正劣于弹性配准的结果。观察者间分割的最大偏差为 15.4 毫米,仿射为 10.5 毫米,弹性为 8.0 毫米。观察者间的平均可变性为 1.5(±1.4)毫米,仿射为 2.8(±2.3)毫米,弹性为 2.2(±1.9)毫米。使用 HERA 的基于强度的弹性配准适用于评估前列腺体积变化,以进行靶区传播和自适应放疗。