Evaluation of the precision and accuracy of augmented reality for pedicle screw placement in the cervical spine

Abstract

Background

Augmented reality (AR) has gained popularity in spine surgery. Head mounted AR devices superimpose a 3D reconstructed model on patient anatomy which has been shown to assist with accurate placement of lumbar spine pedicle screws. We aimed to evaluate the accuracy and precision of AR in cervical spine pedicle screw placement.

Methods

Seven fresh-frozen cadaveric C2-T1 specimens were used. Computed tomography (CT) scans were obtained and uploaded to the AR navigation system. Fiducial markers were utilized to ensure accurate registration. Bilateral C2-C7 pedicle screws were placed. Images containing planned trajectory with tap and navigated screw placement were captured. A post-navigation CT scan was obtained. Time from initial CT scan to navigation and total navigation time were recorded. Radiation dose information was obtained. Linear and angular differences between planned trajectory and navigated screw position as well as navigated screw position and actual screw position were measured on axial and sagittal images. Screw position was graded using the Gertzbein-Robbins classification.

Results

82 pedicle screws were placed. The mean angular and linear deviation between the tap trajectory and navigated screw position were 2.63 ± 2.65° and 3.08 ± 2.32° and 1.11 ± 1.04 mm and 1.24 ± 0.84 mm in the axial and sagittal planes. The mean angular and linear deviation between navigated screw and actual screw were 3.68 ± 4.15° and 2.44 ± 2.17° and 1.51 ± 1.53 mm and 1.02 mm ± 0.88 in the axial and sagittal planes. 95% of screws were given a grade of A or B. Average time from CT scan to navigation was 139.4 seconds and average navigation time was 33 minutes and 46 seconds. Average radiation exposure time was 12.76 ± 1.57 seconds and the average dose-length product (DLP) was 551.15 ± 74.04 mGy-cm.

Conclusions

AR can assist in accurate placement of pedicle screws in the cervical spine. Deviation from navigated screw position to actual screw position was within clinically acceptable range throughout the cervical spine.