Zhao Wei, Brunner Stephen, Niu Kai, Schafer Sebastian, Royalty Kevin, Chen Guang-Hong
Department of Medical Physics, University of Wisconsin-Madison, 1111 Highland Avenue, Madison, WI 53705, USA.
Phys Med Biol. 2015 Feb 7;60(3):1339-65. doi: 10.1088/0031-9155/60/3/1339. Epub 2015 Jan 16.
A patient-specific scatter correction algorithm is proposed to mitigate scatter artefacts in cone-beam CT (CBCT). The approach belongs to the category of convolution-based methods in which a scatter potential function is convolved with a convolution kernel to estimate the scatter profile. A key step in this method is to determine the free parameters introduced in both scatter potential and convolution kernel using a so-called calibration process, which is to seek for the optimal parameters such that the models for both scatter potential and convolution kernel is able to optimally fit the previously known coarse estimates of scatter profiles of the image object. Both direct measurements and Monte Carlo (MC) simulations have been proposed by other investigators to achieve the aforementioned rough estimates. In the present paper, a novel method has been proposed and validated to generate the needed coarse scatter profile for parameter calibration in the convolution method. The method is based upon an image segmentation of the scatter contaminated CBCT image volume, followed by a reprojection of the segmented image volume using a given x-ray spectrum. The reprojected data is subtracted from the scatter contaminated projection data to generate a coarse estimate of the needed scatter profile used in parameter calibration. The method was qualitatively and quantitatively evaluated using numerical simulations and experimental CBCT data acquired on a clinical CBCT imaging system. Results show that the proposed algorithm can significantly reduce scatter artefacts and recover the correct CT number. Numerical simulation results show the method is patient specific, can accurately estimate the scatter, and is robust with respect to segmentation procedure. For experimental and in vivo human data, the results show the CT number can be successfully recovered and anatomical structure visibility can be significantly improved.
提出了一种针对特定患者的散射校正算法，以减轻锥束CT（CBCT）中的散射伪影。该方法属于基于卷积的方法类别，其中散射势函数与卷积核进行卷积以估计散射分布。此方法的关键步骤是使用所谓的校准过程来确定散射势和卷积核中引入的自由参数，即寻找最优参数，以使散射势和卷积核的模型能够最佳地拟合图像对象散射分布的先前已知粗略估计值。其他研究人员已提出直接测量和蒙特卡罗（MC）模拟来实现上述粗略估计。在本文中，提出并验证了一种新颖的方法，用于生成卷积方法中参数校准所需的粗略散射分布。该方法基于对受散射污染的CBCT图像体积进行图像分割，然后使用给定的X射线光谱对分割后的图像体积进行重投影。从受散射污染的投影数据中减去重投影数据，以生成参数校准中所需散射分布的粗略估计值。使用数值模拟和在临床CBCT成像系统上获取的实验CBCT数据对该方法进行了定性和定量评估。结果表明，所提出的算法可以显著减少散射伪影并恢复正确的CT值。数值模拟结果表明，该方法是针对特定患者的，可以准确估计散射，并且对分割过程具有鲁棒性。对于实验数据和体内人体数据，结果表明CT值可以成功恢复，并且解剖结构的可见性可以显著提高。
