Fahrig R, Moreau M, Holdsworth D W
Department of Medical Biophysics, University of Western Ontario, London, Canada.
Med Phys. 1997 Jul;24(7):1097-106. doi: 10.1118/1.598013.
X-ray image intensifiers (XRIIs) have many applications in diagnostic imaging including acquisition of near-real-time projection images of the intracranial and coronary vasculature. Recently, there has been some interest in using this projection data to generate three-dimensional (3-D) computed tomographic (CT) reconstructions. The XRII and x-ray tube are rotated around the object, acquiring sufficient data for the simultaneous reconstruction of many transverse slices. Three-dimensional reconstructions are compromised, however, if the projection data is geometrically distorted in any way. Previous studies have shown the distortion in XRIIs to be substantial and to be highly angular dependent. In this paper, we present a global correction technique which provides a table of correction coefficients for an image acquired at any arbitrary angle about the patient. The coefficients are generated using a linear least-squares fit between the detected and known locations of a grid of small steel beads which is attached to the XRII (27 cm nominal diameter). We have performed corrections on 100 images obtained during rotation of the gantry through 200 degrees and find that a fifth-order polynomial provides optimum image distortion reduction (mean residual distortion of 0.07 pixels), however, fourth-order polynomials provide sufficient distortion reduction for our application (mean residual displacement of 0.1 pixels). Using sixth-order polynomials does not provide a statistically significant reduction in image distortion. The spatial distribution of residual distortion did not demonstrate any particular pattern over the face of the XRII. Image angle and coefficient angle must be known to within +/- 2 degrees in order to keep the mean residual distortion be approximately 0.5 pixels.
X射线影像增强器（XRIIs）在诊断成像中有许多应用，包括获取颅内和冠状动脉血管系统的近实时投影图像。最近，人们对利用这些投影数据生成三维（3-D）计算机断层扫描（CT）重建图像产生了一些兴趣。XRII和X射线管围绕物体旋转，获取足够的数据以同时重建多个横向切片。然而，如果投影数据在任何方面存在几何失真，三维重建就会受到影响。先前的研究表明，XRIIs中的失真很大，并且高度依赖角度。在本文中，我们提出了一种全局校正技术，该技术为围绕患者在任意角度获取的图像提供校正系数表。这些系数是通过对附着在XRII（标称直径27厘米）上的小钢珠网格的检测位置和已知位置之间进行线性最小二乘拟合生成的。我们对在机架旋转200度期间获得的100幅图像进行了校正，发现五次多项式可实现最佳的图像失真减少（平均残余失真为0.07像素），然而，四次多项式对于我们的应用提供了足够的失真减少（平均残余位移为0.1像素）。使用六次多项式并没有在图像失真方面提供统计学上显著的减少。残余失真的空间分布在XRII表面上没有显示出任何特定模式。为了使平均残余失真约为0.5像素，图像角度和系数角度必须精确到±2度以内。
