Computer-assisted navigation is used by surgeons in spine procedures to guide pedicle screws to improve placement
accuracy and in some cases, to better visualize patient’s underlying anatomy. Intraoperative registration is performed to
establish a correlation between patient’s anatomy and the pre/intra-operative image. Current algorithms rely on seeding
points obtained directly from the exposed spinal surface to achieve clinically acceptable registration accuracy. Registration
of these three dimensional surface point-clouds are prone to various systematic errors. The goal of this study was to
evaluate the robustness of surgical navigation systems by looking at the relationship between the optical density of an
acquired 3D point-cloud and the corresponding surgical navigation error. A retrospective review of a total of 48
registrations performed using an experimental structured light navigation system developed within our lab was conducted.
For each registration, the number of points in the acquired point cloud was evaluated relative to whether the registration
was acceptable, the corresponding system reported error and target registration error. It was demonstrated that the number
of points in the point cloud neither correlates with the acceptance/rejection of a registration or the system reported error.
However, a negative correlation was observed between the number of the points in the point-cloud and the corresponding
sagittal angular error. Thus, system reported total registration points and accuracy are insufficient to gauge the accuracy
of a navigation system and the operating surgeon must verify and validate registration based on anatomical landmarks
prior to commencing surgery.
|