A Marker-Free 2D Image-Guided Method for Robot-Assisted Fracture Reduction Surgery

The closed fracture reduction is the first step of the minimally invasive treatment for long bone fractures, while it remains a challenging task due to heavy physical burden, excessive radiation exposure and poor accuracy. The robot-assisted fracture reduction (RAFR) shows potential to address these problems, while most existing methods rely on 3D images, optical tracking equipment and invasive robot-bone fixations. Such methods usually have a tedious surgery workflow, complicated hardware requirements, and iatrogenic injury risks, which prevent them from clinical applications. In this paper, we proposed a novel marker-free RAFR method using the image-based visual servoing technique. Firstly, an image feature vector based on 2D intraoperative images are defined for path planning and feedback control, and an online estimated Jacobian matrix is used to map the image variation to robot motion. Secondly, a geometry-based method is developed for Jacobian initialization without exploratory movements. Thirdly, a novel state-dependent weighted least square (SD-WLS) algorithm for online Jacobian estimation is proposed according to the characteristic of fracture reduction tasks. In the proposed RAFR framework, requirements for 3D images, optical markers and rigid robot-bone fixations are avoided. Experiments on fracture models with highly flexible robot-bone fixations are performed. All the tested fracture cases were well reduced within acceptable steps, which shows the feasibility of the proposed method.