Institute of Medical Physics, University of Erlangen-Nürnberg, Germany.
Phys Med Biol. 2010 Jan 7;55(1):99-120. doi: 10.1088/0031-9155/55/1/007.
Scattered radiation is a major source of artifacts in flat detector computed tomography (FDCT) due to the increased irradiated volumes. We propose a fast projection-based algorithm for correction of scatter artifacts. The presented algorithm combines a convolution method to determine the spatial distribution of the scatter intensity distribution with an object-size-dependent scaling of the scatter intensity distributions using a priori information generated by Monte Carlo simulations. A projection-based (PBSE) and an image-based (IBSE) strategy for size estimation of the scanned object are presented. Both strategies provide good correction and comparable results; the faster PBSE strategy is recommended. Even with such a fast and simple algorithm that in the PBSE variant does not rely on reconstructed volumes or scatter measurements, it is possible to provide a reasonable scatter correction even for truncated scans. For both simulations and measurements, scatter artifacts were significantly reduced and the algorithm showed stable behavior in the z-direction. For simulated voxelized head, hip and thorax phantoms, a figure of merit Q of 0.82, 0.76 and 0.77 was reached, respectively (Q = 0 for uncorrected, Q = 1 for ideal). For a water phantom with 15 cm diameter, for example, a cupping reduction from 10.8% down to 2.1% was achieved. The performance of the correction method has limitations in the case of measurements using non-ideal detectors, intensity calibration, etc. An iterative approach to overcome most of these limitations was proposed. This approach is based on root finding of a cupping metric and may be useful for other scatter correction methods as well. By this optimization, cupping of the measured water phantom was further reduced down to 0.9%. The algorithm was evaluated on a commercial system including truncated and non-homogeneous clinically relevant objects.
散射线是平板探测器 CT（FDCT）中伪影的主要来源，这是由于照射体积的增加。我们提出了一种快速基于投影的算法，用于校正散射伪影。所提出的算法结合了卷积方法来确定散射强度分布的空间分布，以及使用蒙特卡罗模拟生成的先验信息对散射强度分布进行与物体尺寸相关的缩放。提出了一种基于投影（PBSE）和基于图像（IBSE）的扫描物体尺寸估计策略。这两种策略都提供了良好的校正效果和可比的结果；推荐更快的 PBSE 策略。即使是使用这样一种快速而简单的算法，在 PBSE 变体中不依赖于重建体积或散射测量，也可以提供合理的散射校正，即使是对于截断扫描。对于模拟和测量，散射伪影都得到了显著减少，并且算法在 z 方向上表现出稳定的行为。对于模拟的体素化头部、臀部和胸部体模，分别达到了 0.82、0.76 和 0.77 的质量度量 Q（未校正时为 0，理想时为 1）。例如，对于直径为 15 厘米的水模体，杯状伪影从 10.8%减少到 2.1%。在使用非理想探测器、强度校准等情况下，校正方法的性能存在限制。提出了一种迭代方法来克服这些限制中的大多数。这种方法基于杯状度量的根查找，对于其他散射校正方法也可能有用。通过这种优化，测量的水模体的杯状伪影进一步减少到 0.9%。该算法在包括截断和非均匀的临床相关物体的商业系统上进行了评估。
