School of Electrical Engineering, KAIST, Daejeon, Korea.
Department of Radiology, Seoul National University Hospital, Seoul, Korea.
Med Phys. 2017 Nov;44(11):5824-5834. doi: 10.1002/mp.12524. Epub 2017 Sep 13.
A 3D CT image is used along with real-time 2D fluoroscopic images in the state-of-the-art cone-beam CT system to guide percutaneous lung biopsy (PLB). To improve the guiding accuracy by compensating for respiratory motion, we propose an algorithm for real-time matching of 2D fluoroscopic images to multiple 3D CT images of different respiratory phases that is robust to the small movement and deformation due to cardiac motion.
Based on the transformations obtained from nonrigid registration between two 3D CT images acquired at expiratory and inspiratory phases, we first generate sequential 3D CT images (or a 4D CT image) and the corresponding 2D digitally reconstructed radiographs (DRRs) of vessels. We then determine 3D CT images corresponding to each real-time 2D fluoroscopic image, by matching the 2D fluoroscopic image to a 2D DRR.
Quantitative evaluations performed with 20 clinical datasets show that registration errors of anatomical features between a 2D fluoroscopic image and its matched 2D DRR are less than 3 mm on average. Registration errors of a target lesion are determined to be roughly 3 mm on average for 10 datasets.
We propose a real-time matching algorithm to compensate for respiratory motion between a 2D fluoroscopic image and 3D CT images of the lung, regardless of cardiac motion, based on a newly improved matching measure. The proposed algorithm can improve the accuracy of a guiding system for the PLB by providing 3D images precisely registered to 2D fluoroscopic images in real-time, without time-consuming respiratory-gated or cardiac-gated CT images.
在最先进的锥形束 CT 系统中，使用 3D CT 图像和实时 2D 透视图像来指导经皮肺活检（PLB）。为了通过补偿呼吸运动来提高引导准确性，我们提出了一种用于实时匹配 2D 透视图像和多个不同呼吸相位的 3D CT 图像的算法，该算法对由于心脏运动引起的小运动和变形具有鲁棒性。
基于在呼气和吸气阶段采集的两个 3D CT 图像之间的非刚性配准获得的变换，我们首先生成连续的 3D CT 图像（或 4D CT 图像）和相应的血管 2D 数字重建射线照片（DRR）。然后，通过将实时 2D 透视图像与 2D DRR 进行匹配，确定与每个实时 2D 透视图像相对应的 3D CT 图像。
用 20 个临床数据集进行的定量评估表明，2D 透视图像与其匹配的 2D DRR 之间的解剖特征的配准误差平均小于 3mm。对于 10 个数据集，确定目标病变的配准误差平均约为 3mm。
我们提出了一种实时匹配算法，该算法基于新改进的匹配度量标准，补偿了 2D 透视图像与肺部 3D CT 图像之间的呼吸运动，而不受心脏运动的影响。该算法可以通过实时提供与 2D 透视图像精确配准的 3D 图像，来提高 PLB 引导系统的准确性，而无需耗时的呼吸门控或心脏门控 CT 图像。
