基于三维数字图像相关的颅颌面术中骨块位姿自动跟踪关键技术研究

Image－guided surgical navigation has been increasingl popular in craniomaxillofacial surgery, including fracture reduction, orthognathic surgery and maxillofacial reconstruction. Commercially available systems are based on negative infrared tracking technique. The system works by means of detecting the location of probe tip and displaying the point on preoperative radiology images. However, there are two main disadvantages of current navigation technique. Firstly, the target bone segments are three dimensional, but the tracing is based on point, so the system is unable to directly trace the pose and orientation of bone segment. Secondly，the navigation system can’t work automatically , continuous manually interaction using probe is required during tracing process. Such disadvantages increase time expense, decrease the operation fluency and cause inaccuracy.. The goal of present study is to develop a novel six degree-of-freedom automatic tracing system (6DoF-ATS) based on three dimensional-digital imaging correlation (3D-DIC). The other goal is to evaluate the accuracy and convenience of the system. Our research plan consists of the following elements:.(1) Imaging bone segments during operation by means of binocular stereo imaging and 3D-DIC algorithm. .(2) Increase the real-time response and robustness of 3D-DIC algorithm by means of decrease redundancy and GPU parallel calculation..(3) Developing a prototype system and performing validation study to compare the accuracy of 6DoF-ATS and commercial navigation system. . The proposed navigation module in this project will be more convenient, accurate and less time consuming than current navigation systems. It could also be applied in neurosurgery, general surgery and ENT surgery. Moreover, 6DoF-ATS navigation system could be further applied for the control and navigation of surgical assisting robotic system.

在颅颌面手术中骨块就位操作决定术后效果,仅靠医生经验难以做到精确，图像引导技术能够一定程度提高骨块就位的精确性，应用日益广泛。但目前的导航系统无法显示骨块的三维形貌及其位置、姿态，只能进行“点”跟踪，且跟踪过程不能自动完成，需要医生使用探针手动交互，操作复杂，精度不稳定，手术时间长。. 针对上述问题，本项目将对颅颌面手术环境下的双目立体成像和三维数字图像相关技术展开研究，突破算法实时性、鲁棒性两个技术瓶颈，实现颅颌面术中骨块自动位姿跟踪，建立一种新的图像引导模式。拟通过基于纹理的双目立体成像实现骨块的形貌获取和自动跟踪；通过降低算法冗余度和GPU并行计算提高系统的实时性；通过降噪、滤波、差值计算等手段提高算法的鲁棒性；通过医学实验测试系统的引导精度。. 相关成果不仅可推广至骨科等其它专业的导航手术中，还可作为手术辅助机器人的精准作业提供理论和实验基础。

骨块复位或重新就位是颅颌面手术中的一项常见操作，其就位精度直接决定术后效果。近年来，三维图像导航技术开始广泛应用于临床，取代了传统手术依靠医生经验定位的方式，大大提高了手术精度。然而，现有的商业化的导航设备均采用红外定位跟踪的方式，虽然其系统鲁棒性及实时性较好，但由于其只能识别红外标记点，因此在实际应用受到很多限制，其由于其定位方式是通过探针实现的间接定位，导致精度不稳定。在长时间的手术操作过程中，这种交互式的复杂操作方式，有时甚至会成为医生的负累。.针对上述问题，本项目在颅颌面手术环境下，建立了一套基于双目立体视觉和三维数字图像向相关算法的三维导航系统。主要工作和创新点包括：通过降低算法冗余度和GPU并行计算的方式，提高了系统的实时性；通过降噪、滤波、差值计算的方法，提高了算法的鲁棒性；并针对术前选点误差不稳定的问题，提出了一种三维术中模糊配准的改进算法。.综上，本项目实现了颅颌面手术中骨块自动位姿跟踪，并制定了一套新的更适用于临床的图像导航操作流程。相关成果可以推广至导航手术的各个临床领域，并可以作为辅助手术机器人系统的重要组成部分。