Real-time navigation guidance with intraoperative CT imaging for pedicle screw placement using an augmented reality head-mounted display: a proof-of-concept study.

ABSTRACT

OBJECTIVE: Augmented reality (AR) has the potential to improve the accuracy and efficiency of instrumentation placement in spinal fusion surgery, increasing patient safety and outcomes, optimizing ergonomics in the surgical suite, and ultimately lowering procedural costs. The authors sought to describe the use of a commercial prototype Spine AR platform (SpineAR) that provides a commercial AR head-mounted display (ARHMD) user interface for navigation-guided spine surgery incorporating real-time navigation images from intraoperative imaging with a 3D-reconstructed model in the surgeon's field of view, and to assess screw placement accuracy via this method. METHODS: Pedicle screw placement accuracy was assessed and compared with literature-reported data of the freehand (FH) technique. Accuracy with SpineAR was also compared between participants of varying spine surgical experience. Eleven operators without prior experience with AR-assisted pedicle screw placement took part in the study 5 attending neurosurgeons and 6 trainees (1 neurosurgical fellow, 1 senior orthopedic resident, 3 neurosurgical residents, and 1 medical student). Commercially available 3D-printed lumbar spine models were utilized as surrogates of human anatomy. Among the operators, a total of 192 screws were instrumented bilaterally from L2-5 using SpineAR in 24 lumbar spine models. All but one trainee also inserted 8 screws using the FH method. In addition to accuracy scoring using the Gertzbein-Robbins grading scale, axial trajectory was assessed, and user feedback on experience with SpineAR was collected. RESULTS: Based on the Gertzbein-Robbins grading scale, the overall screw placement accuracy using SpineAR among all users was 98.4% (192 screws). Accuracy for attendings and trainees was 99.1% (112 screws) and 97.5% (80 screws), respectively. Accuracy rates were higher compared with literature-reported lumbar screw placement accuracy using FH for attendings (99.1% vs 94.32%; p = 0.0212) and all users (98.4% vs 94.32%; p = 0.0099). The percentage of total inserted screws with a minimum of 5° medial angulation was 100%. No differences were observed between attendings and trainees or between the two methods. User feedback on SpineAR was generally positive. CONCLUSIONS: Screw placement was feasible and accurate using SpineAR, an ARHMD platform with real-time navigation guidance that provided a favorable surgeon-user experience.