Interdisciplinary Center for Scientific Computing, University of Heidelberg, Germany.
Med Phys. 2007 Jul;34(7):3077-92. doi: 10.1118/1.2746486.
This article presents a dynamic CT reconstruction algorithm for objects with time dependent attenuation coefficient. Projection data acquired over several rotations are interpreted as samples of a continuous signal. Based on this idea, a temporal interpolation approach is proposed which provides the maximum temporal resolution for a given rotational speed of the CT scanner. Interpolation is performed using polynomial splines. The algorithm can be adapted to slow signals, reducing the amount of data acquired and the computational cost. A theoretical analysis of the approximations made by the algorithm is provided. In simulation studies, the temporal interpolation approach is compared with three other dynamic reconstruction algorithms based on linear regression, linear interpolation, and generalized Parker weighting. The presented algorithm exhibits the highest temporal resolution for a given sampling interval. Hence, our approach needs less input data to achieve a certain quality in the reconstruction than the other algorithms discussed or, equivalently, less x-ray exposure and computational complexity. The proposed algorithm additionally allows the possibility of using slow rotating scanners for perfusion imaging purposes.
本文提出了一种用于随时间变化衰减系数物体的动态 CT 重建算法。在几个旋转过程中获取的投影数据被解释为连续信号的样本。基于这个想法，提出了一种时间内插方法，为给定的 CT 扫描仪旋转速度提供最大的时间分辨率。内插使用多项式样条。该算法可以适应缓慢的信号，减少采集的数据量和计算成本。提供了对算法所作近似的理论分析。在模拟研究中，将时间内插方法与其他三种基于线性回归、线性内插和广义 Parker 加权的动态重建算法进行了比较。所提出的算法在给定的采样间隔下表现出最高的时间分辨率。因此，与讨论的其他算法相比，我们的方法需要较少的输入数据来达到一定的重建质量，或者等效地说，需要较少的 X 射线曝光和计算复杂度。所提出的算法还允许使用旋转缓慢的扫描仪进行灌注成像。
