Computer Vision-Based Navigation System for High-Precision Orthopedic Trauma Surgery

项目总结/摘要 闭合或开放骨折复位内固定是治疗骨盆骨折的标准手术方法， 目前的临床实践使用荧光镜引导、导丝插入和空心螺钉放置。的 在2D透视中计算复杂3D形态的挑战是不确定性的主要来源，试验和 错误和不良结局，20-30%的螺钉放置不佳和透视时间长（平均透视时间 时间> 123 s），使操作人员暴露于高水平辐射。尽管存在这些挑战， 35年来，手术方法基本上保持不变，手术导航系统（尽管越来越多 在神经外科中常见）存在成本和工作流程障碍，这限制了它们在创伤外科中的广泛应用。 我们提出了一种基于计算机视觉的导航方法，该方法与常规创伤手术工作流程兼容， 提供与立体定向导航相当的准确性的实时引导，使辐射减少十倍 暴露，并与创伤外科武器库中常见的工具一起工作。建议的系统使用一个微型 安装在手术钻上的立体摄像机与荧光透视视图的3D-2D配准相结合， 相对于患者解剖结构的器械轨迹的直接、实时配准。仪器实时叠加 荧光透视视图和/或CT中的轨迹允许准确识别导丝进入点、方向和 骨通道内的一致性，并将减少对“荧光狩猎”和试错导丝的依赖 安置以下目的是开发和评价该系统在骨盆创伤手术中的应用，包括 在临床前研究中对准确性、工作流程和辐射剂量进行定量评估。 目标1。基于计算机视觉的创伤手术引导系统。的硬件和软件组件 将开发一种标准的手术钻，提供实时轨迹， 透视和/或术前CT中的叠加。提出了一种自动钻轴的快速标定方法 校准视频和荧光透视帧的自动基于特征的配准能够实时叠加 X线透视视图中的器械轨迹（X线透视导航）以及CT和X线透视之间的3D-2D配准将 在CT（CT导航）中实现仪器轨迹的实时叠加。 目的2：临床前研究中的评价。基于视觉的导航系统将在临床前 （尸体）实验，以评估准确性和工作流程。这些研究将评价几何精度， 与重复插入尝试次数、手术时间和辐射剂量相关的工作流程因素，评价 与传统徒手透视引导相比，基于视觉的透视导航和CT导航。 该目标的成功实现将建立一个适合于计算机视觉导航的待翻译系统 临床研究在未来的工作。这样的系统为常规创伤手术提供了潜在的重大进步， 功能与最先进的立体定向导航相当，无需成本、复杂性和额外的工作流程 传统的导航。

PROJECT SUMMARY / ABSTRACT Closed or open fracture reduction and internal fixation is the standard surgical approach in treating pelvic fractures, with current clinical practice using fluoroscopic guidance, guidewire insertion, and cannulated screw placement. The challenge in reckoning complex 3D morphology in 2D fluoroscopy presents a major source of uncertainty, trial-and- error, and poor outcomes, with 20-30% rate of suboptimal screw placement and long fluoroscopic runtime (mean fluoro time > 123 s) exposing operating personnel to high levels of radiation exposure. Despite these challenges, mainstream surgical approach has remained largely unchanged for 35 years, and surgical navigation systems (though increasingly common in neurosurgery) present cost and workflow barriers that limit their broad applicability in trauma surgery. We propose a computer vision-based navigation approach that is compatible with routine trauma surgery workflow, offers real-time guidance with accuracy comparable to stereotactic navigation, gives ten-fold reduction in radiation exposure, and works with tools already common in the trauma surgery arsenal. The proposed system uses a miniature stereoscopic camera mounted onboard the surgical drill in combination with 3D-2D registration of fluoroscopic views for direct, real-time registration of the instrument trajectory relative to patient anatomy. Real-time overlay of instrument trajectory in fluoroscopic views and/or CT permits accurate identification of guidewire entry point, orientation, and conformance within bone corridors and will reduce reliance on “fluoro hunting” and trial-and-error guidewire placement. The following aims develop and evaluate the system for application in pelvic trauma surgery, including quantitative assessment of accuracy, workflow, and radiation dose in pre-clinical studies. Aim 1. System for computer vision-based guidance in trauma surgery. The hardware and software components required for vision-based tracking onboard a standard surgical drill will be developed, providing real-time trajectory overlay in fluoroscopy and/or preoperative CT. A fast calibration method will be developed for automatic drill axis calibration. Automatic feature-based registration of the video and fluoroscopic frames enables real-time overlay of instrument trajectory in fluoroscopic views (Fluoro Navigation), and 3D-2D registration between CT and fluoroscopy will enable real-time overlay of the instrument trajectory in CT (CT Navigation). Aim 2: Evaluation in preclinical studies. The vision-based navigation system will be implemented in pre-clinical (cadaver) experiments to evaluate accuracy and workflow. These studies will evaluate the geometric accuracy and workflow factors relating to the number of repeated insertion attempts, procedure time, and radiation dose, evaluating vision-based Fluoro Navigation and CT Navigation in comparison to conventional freehand fluoroscopy guidance. Successful completion of the aims will establish a system suitable for computer vision-based navigation to be translated to clinical studies in future work. Such a system offers a potentially major advance in routine trauma surgery, bringing capabilities comparable to state-of-the-art stereotactic navigation without the cost, complexity, and additional workflow of conventional navigation.