Development and Evaluation of CT-to-3D Ultrasound Image Registration Algorithm in Vertebral Phantoms for Spine Surgery

Posterior spinal fusion surgery requires careful insertion of screws into the spine to avoid neurologic injury. While current systems use CT-scans, three-dimensional ultrasound (3DUS) could provide guidance by reconstructing the vertebral surface, and then registering a pre-operative vertebral model to that surface for localization. The aim of this study was to evaluate the accuracy and processing time of a custom CT-3DUS registration algorithm. A phantom human vertebra was 3D-printed and scanned with a motion capture-based 3D ultrasound (3DUS) system. Image registration was performed that included a pre-alignment phase using vertebral symmetry information, and then comparing Gaussian pyramid intensity-based registration with iterative-closest-point registration for final transformations. Image registration was performed 192 times while surgical registration between CT and real-world position was performed 84 times. The accuracy of image registration (CT-to-3DUS) was 0.3 +/- 0.2 mm and 0.9 +/- 0.8 degrees completed in 13.3 +/- 2.9 s. The surgical navigation accuracy (CT model to real-world position) of the system was 1.2 +/- 0.5 mm and 2.2 +/- 2.0 degrees completed in 16.2 +/- 3.0 s. Both meet accuracy thresholds of < 2 mm and < 5 degrees required for the surgery. A feasibility study on porcine spine qualitatively showed appropriate overlapping anatomy in CT-3DUS registrations. The usage of 3D ultrasound for navigation has demonstrated accuracy to provide radiation-free image guidance for spine surgery.