Auto-registration is better than point-to-point registration with respect to clinical accuracy when using the same active infrared navigation system during spinal surgery. This is the conclusion of the first study to compare the difference between the two methods with respect to clinical accuracy, originally published in Spine.
Computer assisted navigation technology has been widely used in spinal surgery since the early 21st century. There are two major registration methods in the navigation system: point-to-point registration and auto-registration. Point-to-point registration is based on preoperative CT images and requires point-to point matching during surgery. Auto-registration is based on an intraoperative cone beam CT scan, and does not require point matching during surgery.
The authors of the study note that, since the mechanisms of the two registration methods are different, the accuracy varies. The maximum permissible translational error of a pedicle screw at the cervical spine, the mid-thoracic spine, and the thoracolumbar junction is less than 1mm, as evidenced in the literature.
Clinical accuracy in this experiment was defined as the distance between the navigation coordinate and the image coordinate of the same point. The study authors say, “The navigation coordinate and the image coordinate represented the positions of the same point in the coordinate frames of the navigation system and the CT image respectively. Thus, the deviation between the two coordinates can represent the accuracy of the navigation system.”
The clinical accuracy of the point-to-point registration group varied significantly among different segments. The average accuracy of the most accurate segment, L2, which provided the reference points during the point-topoint registration, was 1.10 (standard deviation 0.61mm). In the two segments adjacent to L2, the clinical accuracy was found to be significantly worse than in L2—in fact, the further the segments from L2, the worse the clinical accuracy, with the accuracy of L4 and L5 over 2mm. This is due to the varying flexibility of the spine along its length.
The clinical accuracies described for point-to-point registration is sufficient for surgeries in low risk segments, such as the lower lumbar. However, the authors note, “The accuracy of the reference segment can barely meet the requirement of surgeries in cervical, midthoracic or thoracolumbar junction spine.”
In contrast, in the auto-registration group, there was no significant difference in the clinical accuracy found among different segments. The average accuracy of all segments was 0.74mm (standard deviation 0.3), which is significantly better than in the most accurate segment in point-to-point registration. This clinical accuracy is sufficient for surgeries in high-risk segments of the spinal column.
The investigators summarise: “In comparison [to point-to-point registration], the accuracy of auto-registration is relevant to neither the image clarity nor the human finesse, which makes it more safe and feasible in difficult cases. Further studies based on real setups are required to investigate the efficacy of the surface matching.”
To conduct this study, Wei Tian and Yajun Liu (both Department of Spine Surgery, Beijing Jishuitan Hospital; Beijing Key Laboratory of Robotic Orthopaedics, Beijing, China) and colleagues used a novel method to measure the clinical accuracy of the two navigation systems. They worked on an operative set that mimicked a regular surgery in all ways, except that the patient was replaced by a 1:1 ratio model, made out of Sawbone. Surgeon gloves, the lamps, and the draping of the surgical field all exactly replicated a normal operating room setup. Titanium beads were placed firmly onto different parts of the bone model, and could be accurately located in CT images. The surgeons used the navigation system Nav3, with Stryker’s SpineMap 3D software.