Zhang Bin, Zeng Gengsheng L
Department of Electrical and Computer Engineering, University of Utah, Salt Lake City, UT 84112, USA.
Phys Med Biol. 2006 Nov 21;51(22):5825-42. doi: 10.1088/0031-9155/51/22/007. Epub 2006 Oct 24.
Spherically symmetric volume elements (blobs) have better resolution-noise performance than voxels because of the overlapping of their rotational symmetric basis functions; however, using blobs is more computationally expensive than using voxels due to blob overlap. In this paper, we propose an immediate after-backprojection filtering method (ABF) with blob-shaped window functions for a voxel-based reconstruction. We compared this method with the general voxel-based method (without filtering), the blob-based method, the voxel-based method with between-iteration filtering (BIF) and with post-filtering (POF), using computer simulations. Both the quality of the reconstruction and the computational cost were evaluated. The reconstruction quality was measured by the contrast recovery coefficient (CRC) versus the background noise. It is shown that images reconstructed using this method are characterized by less image noise and preserved image contrast in comparison with both the general voxel-based method and the voxel-based method with BIF. The improvement in image quality achieved by this method varies with the parameters chosen for the Kaiser-Bessel (KB) windows. As with blobs, wider KB windows achieve better contrast-noise trade-offs in the reconstructed images, but are more computationally expensive. When using a KB window of a = 2.0, alpha = 10.4 and m = 2, known as the basis function of a 'standard' blob, this new method achieves identical CRC-noise features to the blob-based method with 'standard' blobs. In addition, the ABF method can be combined with the post-filtering method to achieve better noise-resolution performance than the general voxel-based post-filtering method. The computational cost of the ABF method is slightly greater than that of the general voxel-based method, but much less than that of the blob-based method.
由于其旋转对称基函数的重叠，球形对称体积元素（团块）比体素具有更好的分辨率-噪声性能；然而，由于团块重叠，使用团块比使用体素在计算上更昂贵。在本文中，我们提出了一种基于体素重建的、带有团块形状窗口函数的直接反投影滤波方法（ABF）。我们使用计算机模拟将该方法与普通的基于体素的方法（无滤波）、基于团块的方法、带有迭代间滤波（BIF）和后滤波（POF）的基于体素的方法进行了比较。评估了重建质量和计算成本。通过对比恢复系数（CRC）与背景噪声来衡量重建质量。结果表明，与普通的基于体素的方法和带有BIF的基于体素的方法相比，使用该方法重建的图像具有更少的图像噪声和保留的图像对比度。该方法实现的图像质量提升随为凯泽-贝塞尔（KB）窗口选择的参数而变化。与团块一样，更宽的KB窗口在重建图像中实现了更好的对比度-噪声权衡，但计算成本更高。当使用α = 2.0、α = 10.4和m = 2的KB窗口（称为“标准”团块的基函数）时，这种新方法实现了与基于“标准”团块的方法相同的CRC-噪声特征。此外，ABF方法可以与后滤波方法相结合，以实现比普通基于体素的后滤波方法更好的噪声分辨率性能。ABF方法的计算成本略高于普通基于体素的方法，但远低于基于团块的方法。
