Nuclear and Radiological Engineering and Medical Physics Programs, The George W. Woodruff School of Mechanical Engineering, Georgia Institute of Technology, Atlanta, GA, 30332, USA.
Landauer Medical Physics, 2 Science Road, Glenwood, IL, 60425, USA.
Med Phys. 2017 May;44(5):1796-1808. doi: 10.1002/mp.12190. Epub 2017 Apr 17.
The image quality of cone beam computed tomography (CBCT) is limited by severe shading artifacts, hindering its quantitative applications in radiation therapy. In this work, we propose an image-domain shading correction method using planning CT (pCT) as prior information which is highly adaptive to clinical environment.
We propose to perform shading correction via sparse sampling on pCT. The method starts with a coarse mapping between the first-pass CBCT images obtained from the Varian TrueBeam system and the pCT. The scatter correction method embedded in the Varian commercial software removes some image errors but the CBCT images still contain severe shading artifacts. The difference images between the mapped pCT and the CBCT are considered as shading errors, but only sparse shading samples are selected for correction using empirical constraints to avoid carrying over false information from pCT. A Fourier-Transform-based technique, referred to as local filtration, is proposed to efficiently process the sparse data for effective shading correction. The performance of the proposed method is evaluated on one anthropomorphic pelvis phantom and 17 patients, who were scheduled for radiation therapy. (The codes of the proposed method and sample data can be downloaded from https://sites.google.com/view/linxicbct) RESULTS: The proposed shading correction substantially improves the CBCT image quality on both the phantom and the patients to a level close to that of the pCT images. On the phantom, the spatial nonuniformity (SNU) difference between CBCT and pCT is reduced from 74 to 1 HU. The root of mean square difference of SNU between CBCT and pCT is reduced from 83 to 10 HU on the pelvis patients, and from 101 to 12 HU on the thorax patients. The robustness of the proposed shading correction is fully investigated with simulated registration errors between CBCT and pCT on the phantom and mis-registration on patients. The sparse sampling scheme of our method successfully avoids false structures in the corrected CBCT even when the maximum registration error is as high as 8 mm.
We develop an effective shading correction algorithm for CBCT readily implementable on clinical data as a software plug-in without modifications of current imaging hardware and protocol. The algorithm is directly applied on the output images from a commercial CBCT scanner with high computational efficiency and negligible memory burden.
锥形束 CT（CBCT）的图像质量受到严重的阴影伪影的限制，阻碍了其在放射治疗中的定量应用。在这项工作中，我们提出了一种基于图像域的阴影校正方法，该方法使用规划 CT（pCT）作为高度适应临床环境的先验信息。
我们提出通过稀疏采样在 pCT 上进行阴影校正。该方法首先在瓦里安 TrueBeam 系统获得的首次通过 CBCT 图像和 pCT 之间建立粗略映射。嵌入在瓦里安商业软件中的散射校正方法去除了一些图像误差，但 CBCT 图像仍然存在严重的阴影伪影。映射的 pCT 和 CBCT 之间的差异图像被认为是阴影误差，但仅选择稀疏的阴影样本进行校正，使用经验约束避免从 pCT 传递错误信息。提出了一种基于傅里叶变换的技术，称为局部滤波，用于有效地处理稀疏数据以进行有效的阴影校正。该方法的性能在一个人体骨盆体模和 17 名接受放射治疗的患者上进行了评估。（该方法的代码和示例数据可以从 https://sites.google.com/view/linxicbct 下载）
该方法显著改善了体模和患者上的 CBCT 图像质量，使其接近 pCT 图像的水平。在体模上，CBCT 和 pCT 之间的空间不均匀性（SNU）差异从 74 减少到 1 HU。骨盆患者的 CBCT 和 pCT 之间的 SNU 均方根差异从 83 减少到 10 HU，胸患者从 101 减少到 12 HU。在体模上模拟 CBCT 和 pCT 之间的注册误差以及患者上的配准错误，充分研究了所提出的阴影校正的鲁棒性。即使最大注册误差高达 8mm，我们的方法的稀疏采样方案也成功避免了校正后的 CBCT 中的虚假结构。
我们开发了一种有效的 CBCT 阴影校正算法，可作为软件插件直接应用于商业 CBCT 扫描仪的输出图像，无需修改当前成像硬件和协议。该算法具有很高的计算效率和可忽略的内存负担。
