Percutaneous placement of lumbar pedicle screws via intraoperative CT image-based augmented reality-guided technology

The authors aimed to assess, in a bone-agar experimental setting, the feasibility and accuracy of percutaneous lumbar pedicle screw placements using an intraoperative CT image-based augmented reality (AR)-guided method compared to placements using a radiograph-guided method. They also compared two AR hologram alignment methods.

Twelve lumbar spine sawbones were completely embedded in hardened opaque agar, and a cubic marker was fixed on each phantom. After intraoperative CT, a 3D model of each phantom was generated, and a specialized application was deployed into an AR headset (Microsoft HoloLens). One hundred twenty pedicle screws, simulated by Kirschner wires (K-wires), were placed by two experienced surgeons, who each placed a total of 60 screws: 20 placed with a radiograph-guided technique, 20 with an AR technique in which the hologram was manually aligned, and 20 with an AR technique in which the hologram was automatically aligned. For each K-wire, the insertion path was expanded to a 6.5-mm diameter to simulate a lumbar pedicle screw. CT imaging of each phantom was performed after all K-wire placements, and the operative time required for each K-wire placement was recorded. An independent radiologist rated all images of K-wire placements. Outcomes were classified as grade I (no pedicle perforation), grade II (screw perforation of the cortex by up to 2 mm), or grade III (screw perforation of the cortex by > 2 mm). In a clinical situation, placements scored as grade I or II would be acceptable and safe for patients.

Among all screw placements, 75 (94%) of 80 AR-guided placements and 40 (100%) of 40 radiograph-guided placements were acceptable (i.e., grade I or II; p = 0.106). Radiograph-guided placements had more grade I outcomes than the AR-guided method (p < 0.0001). The accuracy of the two AR alignment methods (p = 0.526) was not statistically significantly different, and neither was it different between the AR and radiograph groups (p < 0.0001). AR-guided placements required less time than the radiograph-guided placements (mean ± standard deviation, 131.76 ± 24.57 vs 181.43 ± 15.82 seconds, p < 0.0001). Placements performed using the automatic-alignment method required less time than those using the manual-alignment method (124.20 ± 23.80 vs 139.33 ± 23.21 seconds, p = 0.0081).

In bone-agar experimental settings, AR-guided percutaneous lumbar pedicle screw placements were acceptable and more efficient than radiograph-guided placements. In a comparison of the two AR-guided placements, the automatic-alignment method was as accurate as the manual method but more efficient. Because of some limitations, the AR-guided system cannot be recommended in a clinical setting until there is significant improvement of this technology.