Department of Interventional Guidance Technology, Philips Research North America, Briarcliff Manor, New York, NY 10510, USA.
Int J Comput Assist Radiol Surg. 2011 May;6(3):407-19. doi: 10.1007/s11548-010-0518-4. Epub 2010 Aug 17.
Electromagnetic (EM)-guided endoscopy has demonstrated its value in minimally invasive interventions. Accuracy evaluation of the system is of paramount importance to clinical applications. Previously, a number of researchers have reported the results of calibrating the EM-guided endoscope; however, the accumulated errors of an integrated system, which ultimately reflect intra-operative performance, have not been characterized. To fill this vacancy, we propose a novel system to perform this evaluation and use a 3D metric to reflect the intra-operative procedural accuracy.
This paper first presents a portable design and a method for calibration of an electromagnetic (EM)-tracked endoscopy system. An evaluation scheme is then described that uses the calibration results and EM-CT registration to enable real-time data fusion between CT and endoscopic video images. We present quantitative evaluation results for estimating the accuracy of this system using eight internal fiducials as the targets on an anatomical phantom: the error is obtained by comparing the positions of these targets in the CT space, EM space and endoscopy image space. To obtain 3D error estimation, the 3D locations of the targets in the endoscopy image space are reconstructed from stereo views of the EM-tracked monocular endoscope. Thus, the accumulated errors are evaluated in a controlled environment, where the ground truth information is present and systematic performance (including the calibration error) can be assessed.
We obtain the mean in-plane error to be on the order of 2 pixels. To evaluate the data integration performance for virtual navigation, target video-CT registration error (TRE) is measured as the 3D Euclidean distance between the 3D-reconstructed targets of endoscopy video images and the targets identified in CT. The 3D error (TRE) encapsulates EM-CT registration error, EM-tracking error, fiducial localization error, and optical-EM calibration error.
We present in this paper our calibration method and a virtual navigation evaluation system for quantifying the overall errors of the intra-operative data integration. We believe this phantom not only offers us good insights to understand the systematic errors encountered in all phases of an EM-tracked endoscopy procedure but also can provide quality control of laboratory experiments for endoscopic procedures before the experiments are transferred from the laboratory to human subjects.
电磁（EM）引导内窥镜在微创介入中已经显示出了其价值。系统的准确性评估对于临床应用至关重要。此前，许多研究人员已经报告了对 EM 引导内窥镜进行校准的结果；然而，尚未对集成系统的累积误差进行描述，而这些误差最终反映了手术过程中的性能。为了填补这一空白，我们提出了一种新的系统来进行这种评估，并使用 3D 度量来反映手术过程中的程序准确性。
本文首先介绍了一种用于电磁（EM）跟踪内窥镜系统的便携式设计和校准方法。然后描述了一种评估方案，该方案使用校准结果和 EM-CT 配准来实现 CT 和内窥镜视频图像之间的实时数据融合。我们使用解剖学模型上的八个内部基准作为目标，提出了一种使用定量评估结果来估计该系统准确性的方法：通过比较这些目标在 CT 空间、EM 空间和内窥镜图像空间中的位置来获得误差。为了获得 3D 误差估计，从 EM 跟踪单目内窥镜的立体视图重建内窥镜图像空间中目标的 3D 位置。因此，在存在真实信息且可以评估系统性能（包括校准误差）的受控环境中评估累积误差。
我们获得的平面内平均误差约为 2 个像素。为了评估虚拟导航的数据集成性能，测量了目标视频-CT 配准误差（TRE），即内窥镜视频图像的 3D 重建目标与 CT 中识别的目标之间的 3D 欧几里得距离。3D 误差（TRE）包含 EM-CT 配准误差、EM 跟踪误差、基准定位误差和光学-EM 校准误差。
我们在本文中提出了我们的校准方法和虚拟导航评估系统，用于量化手术过程中数据集成的整体误差。我们相信，这种模型不仅为我们提供了很好的洞察力，以了解 EM 跟踪内窥镜手术过程中所有阶段遇到的系统误差，而且还可以为内窥镜手术前从实验室转移到人体实验的实验室实验提供质量控制。
