X-ray Image-Based Pose Estimation of a Joint-Encoded Spinal Surgical Positioning Arm

Minimally invasive spinal surgery requires accurate and efficient surgical tool positioning; however, current optical or navigation-assisted systems can be costly, complex, or expose patients to increased radiation. To address these challenges, we propose and evaluate a novel passive spinal surgical positioning arm with X-ray image-based pose estimation capability. The system estimates the 6-degree-of-freedom pose of the arm using radiographic landmarks extracted from a single intraoperative X-ray image. In our experiment, a surgeon performed single-level pedicle screw placement on three sawbones spine models. The estimated positioning accuracy achieved a mean translational error of 1.08 ± 0.40 mm. The average setup time was 2.3 ± 0.9 minutes, and the surgical operation time was 4.2 ± 1.8 minutes. Only 6.75 ± 0.95 X-ray images were required per procedure, resulting in an average radiation dose of 0.27 ± 0.04 mGy. These results demonstrate that our system can potentially reduce radiation exposure and surgical duration while maintaining high positioning accuracy. Further clinical validation is required to confirm its effectiveness in real-world surgical settings.