Single vertebral registration method for minimally invasive spine surgery robot

Abstract

For minimally invasive spinal surgical robot, preoperative images and intraoperative image registration technology is the core. Image registration is intended to show the trajectory of preoperative planning images into intraoperative image, and also show the relative position of the surgical instruments and trajectory of preoperative planning. As the physiological curve, anatomical structure of the spine and the intervertebral discs and ligaments, the entire spine can be treated as a deformable structure. Simultaneously, the different capture angle of the preoperative images and intraoperative image will also lead to the registration errors. Therefore, this paper proposes a method based on SIFT operator and multi-angle search for a single vertebra. This method is realized in the preoperative and intraoperative image registration on the anteroposterior and lateral view. For the verification of the registration results, we use opre-implantation Kirschner, and the algorithm based on point to point to verify the registration method. As the validation error point-based on Kirschner wire exists a certain space limitations, another verification method, Harris corner error validation method could compensate this limitation. By detecting the Harris corner in preoperative and intraoperative image, we calculate the multi-point based error to validate the registration method, the errorsare all less than 1 mm. It is acceptable for laboratory examination and clinical application.