Beijing Engineering Research Center of Mixed Reality and Advanced Display, School of Optics and Photonics, Beijing Institute of Technology, Beijing, 100081, China.
Institute of Engineering Medicine, Beijing Institute of Technology, Beijing, 100081, China.
Int J Comput Assist Radiol Surg. 2022 Jun;17(6):981-991. doi: 10.1007/s11548-022-02589-0. Epub 2022 Mar 14.
Facial repair surgeries (FRS) require accuracy for navigating the critical anatomy safely and quickly. The purpose of this paper is to develop a method to directly track the position of the patient using video data acquired from the single camera, which can achieve noninvasive, real time, and high positioning accuracy in FRS.
Our method first performs camera calibration and registers the surface segmented from computed tomography to the patient. Then, a two-step constraint algorithm, which includes the feature local constraint and the distance standard deviation constraint, is used to find the optimal feature matching pair quickly. Finally, the movements of the camera and the patient decomposed from the image motion matrix are used to track the camera and the patient, respectively.
The proposed method achieved fusion error RMS of 1.44 ± 0.35, 1.50 ± 0.15, 1.63 ± 0.03 mm in skull phantom, cadaver mandible, and human experiments, respectively. The above errors of the proposed method were lower than those of the optical tracking system-based method. Additionally, the proposed method could process video streams up to 24 frames per second, which can meet the real-time requirements of FRS.
The proposed method does not rely on tracking markers attached to the patient; it could be executed automatically to maintain the correct augmented reality scene and overcome the decrease in positioning accuracy caused by patient movement during surgery.
面部修复手术 (FRS) 需要精确地安全快速地导航关键解剖结构。本文的目的是开发一种方法，该方法可以使用从单摄像机获取的视频数据直接跟踪患者的位置，从而在 FRS 中实现非侵入性、实时和高精度的定位。
我们的方法首先进行摄像机标定，并将从计算机断层扫描中分割出的表面与患者配准。然后，使用两步约束算法，包括特征局部约束和距离标准差约束，快速找到最佳特征匹配对。最后，使用从图像运动矩阵中分解出的摄像机和患者的运动来分别跟踪摄像机和患者。
所提出的方法在颅骨模型、尸体下颌骨和人体实验中的融合误差 RMS 分别达到了 1.44±0.35、1.50±0.15 和 1.63±0.03 毫米。与基于光学跟踪系统的方法相比，该方法的误差更低。此外，该方法可以处理高达 24 帧/秒的视频流，这可以满足 FRS 的实时要求。
所提出的方法不依赖于贴在患者身上的跟踪标记；它可以自动执行，以保持正确的增强现实场景，并克服手术过程中患者运动导致的定位精度下降。
