IEEE Trans Med Imaging. 2019 Apr;38(4):1048-1057. doi: 10.1109/TMI.2018.2877503. Epub 2018 Oct 23.
For wrist complaints related to motion, a 2-D radiograph or CT scan of the static wrist may not always be considered diagnostic. 3-D motion imaging, i.e., multiple 3DCT scans in time (4DCT), enables quantifying carpal motion and comparing motion patterns of the affected wrist with those of the healthy contralateral side. The accuracy and precision of the method, however, is limited by noise and motion artifacts. Although, the technique is considered promising in existing literature, the accuracy and precision of carpal motion analysis has never been investigated systematically. In this paper, we introduce and evaluate a semi-automatic segmentation- and registration-based method for 3-D carpal motion analysis. We investigate the accuracy and precision of the method, and its dependency on motion and scan parameters (angular velocity, dose, gantry revolution angle for image reconstruction, and scanner type) using a wrist phantom. During standstill the positioning error was ≤ 0.23 mm and ≤ 0.78°. A partial gantry revolution for 3-D reconstruction introduced image deformation, contributing to a positioning error of approx. 0.8 mm. This error increased with reduced dose, and with increasing angular velocity of the wrist phantom. In cases where the phantom was rotating about an axis parallel to the rotation axis of the gantry, and in a direction opposite to the gantry, the positioning error increased, probably because of the apparent increase in angular velocity with respect to the gantry. Slow carpal motion 4DCT analysis is feasible using a regular CT scanner. A partial gantry revolution angle for 3-D reconstruction may introduce image deformation, which decreases the accuracy of carpal motion analysis. Knowing the positioning error in 4DCT imaging with the proposed method is considered valuable when investigating wrist injury since it enables discrimination of actual motion from apparent motion caused by methodological error.
对于与运动相关的腕部投诉,静态腕部的二维射线照相或 CT 扫描并不总是被认为具有诊断价值。三维运动成像,即多个时间(4DCT)的三维 CT 扫描,能够量化腕骨运动,并将受影响手腕的运动模式与健康对侧手腕的运动模式进行比较。然而,该方法的准确性和精密度受到噪声和运动伪影的限制。尽管该技术在现有文献中被认为很有前途,但腕骨运动分析的准确性和精密度从未被系统地研究过。在本文中,我们介绍并评估了一种基于半自动分割和配准的三维腕骨运动分析方法。我们使用腕部模型研究了该方法的准确性和精密度,以及其对运动和扫描参数(角速度、剂量、用于图像重建的机架旋转角度和扫描仪类型)的依赖性。在静止状态下,定位误差≤0.23mm 和≤0.78°。三维重建时部分机架旋转会导致图像变形,从而导致定位误差约为 0.8mm。该误差随着剂量的降低以及腕部模型角速度的增加而增加。在模型绕与机架旋转轴平行且与机架旋转方向相反的轴旋转的情况下,定位误差增加,这可能是由于相对于机架角速度的明显增加所致。使用常规 CT 扫描仪可以进行缓慢的腕部 4DCT 分析。三维重建时部分机架旋转角度可能会导致图像变形,从而降低腕骨运动分析的准确性。在研究腕部损伤时,了解所提出方法在 4DCT 成像中的定位误差是有价值的,因为它能够区分实际运动和由于方法误差引起的表观运动。