Accuracy of a high-resolution augmented reality (AR) guidance system with novel 3D stereoscopic targeting for spinal surgery Authors: Muwaffak Abdulhak, Sudesh K. Srivastav, John Pan, Hans-Ulrich Bueff, Thomas Voegeli, Serena S. Hu Published: 2026 ABSTRACT BackgroundAugmented reality (AR) is increasingly transforming the landscape of spine surgery by offering enhanced visualization of anatomical structures in real time, directly overlaid onto the surgical field. The study objective was to determine the accuracy of a novel high-resolution AR system with novel three-dimensional (3D) stereoscopic targeting in pedicle screw insertion and to compare it with published studies on existing AR, navigation, and robotic systems.MethodsEighty pedicle screws were inserted using a novel high-resolution AR system (minimally invasive 55, open 25; thoracic 34, lumbosacral 46). Heary and Gertzbein-Robbins grading was applied, assessing the degree of pedicle breaches on postoperative CT scans. 3D measurements of positional error (PE) and angular error (AE) were performed by comparing post-operative CT scans of implanted screws with planned trajectories.ResultsSeventy-six screws placed had Heary grade I. Four screws had a larger diameter than the pedicle and were planned with an in-out-in technique with an intended Heary grade II. Screw placement accuracy rate defined as the percentage of implantations with Heary grades I and II and Gertzbein-Robbins grades A and B was 100%. PE and AE (mean ± standard deviation) for the novel AR system were 1.1±0.5 mm and 1.3°±0.5°, respectively. Percent increment in PE and AE for different navigation, robotic, and AR systems compared to the novel AR system ranged between 23.7–80.6% and 40.6–79.5%, respectively, with the differences being highly significant for all comparisons.ConclusionsPE and AE for placing pedicle screws are 2–5× smaller for the high-resolution AR system with novel 3D stereoscopic targeting compared to published studies on existing navigation, robotic and AR systems. A PE of 1.1 mm and AE of 1.3° makes this technology particularly suited for small pedicles and challenging anatomy. Highlight box Key findings • Augmented reality (AR) with novel three-dimensional (3D) stereoscopic targeting is highly accurate in pedicle screw placement with a 2–5× improvement in position and angular error (AE) compared to published studies on existing navigation, robotics, and AR systems. What is known and what is new? • AR provides enhanced intuitive and immersive assistance for intra-operative decision-making with improved hand-eye coordination. AR preserves tactile feedback unlike robotic systems. • 3D stereoscopic targeting with visual magnification of positional and AEs has the potential to improve surgical accuracy compared to published studies on existing navigation, robotic, and AR techniques. What is the implication, and what should change now? • Achieve high surgical accuracy with the potential to enhance patient safety. • Novel 3D stereoscopic targeting with use of AR display appears to be particularly well suited for small pedicles and challenging anatomy. Introduction Spinal fusion is one of the most common surgeries in the U.S., with approximately 1.36 million procedures performed in 2021 based on analysis of Current Procedural Terminology (CPT) and International Classification of Diseases (ICD) codes (1). In a series of 4,790 pedicle screw placements, a complication rate of 2.4% was reported per screw (2); the most frequent complication was nerve root irritation and damage from screw angulation through the medial cortex (2). While surgical navigation with computed tomography (CT) can improve the accuracy of screw placement (3), high capital cost (4), in particular when combined with expensive intra-operative cone beam CT scanners (4), can limit its broad applicability. Capital expenditure is of even greater concern with robotic systems for pedicle screw placement (4-6), which can cost more than $1 million upfront cost plus high operating costs. With navigation or robotics, imaging studies of the patient are displayed in the operating room (OR) on an external computer monitor, several feet away from the patient (3,5,7). Since the surgical field is in a different location and has different coordinates than the computer monitor, hand-eye coordination can be challenging for the surgeon. Augmented reality (AR) is increasingly transforming the landscape of spine surgery by offering enhanced visualization of anatomical structures in real time, directly overlaid onto the surgical field (8-12). By integrating imaging studies, such as intra-operative cone-beam CT, with the intraoperative view of the physical spine, AR enables more precise localization of spinal anatomy, potentially improving the accuracy of pedicle screw placement and reducing reliance on fluoroscopy. This technology can enhance spatial awareness without the need to shift attention away from the operative site, thereby streamlining workflow and minimizing cognitive load (8-12). Compared to traditional methods such as freehand techniques or image-guided navigation, AR promises a more intuitive and immersive interface for intraoperative decision-making. There is, however, a paucity of studies reporting on the accuracy of AR in spinal surgery. The current study addresses this issue by assessing both the radiographic (measured via Heary and Gertzbein Robbins classifications) and quantitative accuracy [measured in three-dimensional (3D) with computer software] of a novel AR system for spinal surgery with novel 3D stereoscopic targeting measured in three dimensions, and comparing it with published studies on existing navigation, robotic and AR systems. We present this article in accordance with the TREND reporting checklist (available at https://jss.amegroups.com/article/view/10.21037/jss-2026-1-0034/rc). Methods Surgeons, sample size Screws were planned and placed by two neurosurgeons and one orthopedic surgeon experienced in spine surgery and computer-aided surgery, including navigation and robotics. Each surgeon had more than 10 years of experience in spine surgery. Surgeons would work in tandem, with each surgeon implanting one side of the patient. The surgeons had not utilized the novel AR system prior to the cadaver surgeries; they were given a brief introduction and sawbones demonstration of the AR guidance system before starting cadaver tests. Pedicle screw placement accuracy was tested in five anonymized cadaver specimens sourced from Science Care (Phoenix, AZ, USA), following their compliance and safety protocols accredited by the Association for Advancing Tissue and Biologics. Three of the cadavers were osteoporotic with one or more vertebral compression fractures. A total of 80 pedicle screws were implanted. Fifty-five screws were implanted using a minimally invasive technique. Twenty-five screws were implanted using open exposure. Cadaver specimens were alternated between MIS and open techniques. There were 34 thoracic screws implanted (T5–12) and 46 lumbosacral screws (L1–S1). All 80 screws were implanted using intra-operative imaging with a cone beam CT scanner (O-arm, Medtronic, Minneapolis, MN, USA). The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. Since cadaver specimens were anonymized and donor identity was not available to the investigators, no ethics committee approval was required. Screw systems used Fifty-eight Medtronic Solera Voyager screws (Medtronic, Minneapolis, MN, USA), 10 Stryker Everest MIS screws (Stryker, Kalamazoo, MI, USA), 8 Globus Creo screws (Globus Medical, Inc., Audubon, PA, USA); 2 Johnson & Johnson DePuy Viper Prime screws (J&J DePuy Synthes, Raynham, MA, USA), and 2 Alphatec Invictus MIS screws (Alphatec Holdings, Carlsbad, CA, USA) were available for implantation. The fiducial array adapters of the novel AR system are implant company agnostic and are configured to work across all major screw and interbody device implant systems. When the surgeon encountered a small pedicle and the available screw diameter exceeded the pedicle width, these screws were planned using an in-out-in technique, and a small lateral breach was determined acceptable in the area of concavity between the pedicle and the posterior aspect of the vertebral body. AR guidance system The novel AR guidance system (OnPoint AR-AI, OnPoint Surgical, Inc., Concord, MA, USA) is provided to the OR with a complete navigation system on a small cart comprising a tracking camera and two computer monitors. The navigation system uses infrared tracking for tracking fiducial arrays attached to pelvic pins or spinal clamps, surgical instruments, and the AR headsets (Figure 1A). The novel AR system can be used with open surgery or minimally invasive surgery (MIS). The AR guidance system is configured to track pedicle screw systems, interbody device systems, or other systems or instruments that use trajectory planning. The AR guidance system is provided with a set of array adapters that can be attached to rotating instruments or instruments with fixed, non-rotating components. The instrument array adapter system is compatible with manual surgery and power drills (Figure 1A). Surgical steps using AR technique. (A) AR guidance of SD using F with attached universal RA. The surgeon is using a PD (Powerease, Medtronic, Minneapolis, MN, USA). The head-mounted display includes optional L and HL. (B) Example of AR display provided by the novel AR system. A standard navigation display is seen on the right with a pre-planned screw (yellow) and a tracked virtual instrument (thick light blue line) and the instrument trajectory (thin light blue line). On the left, alphanumeric numbers are provided: 4.7 mm positional deviation, 1.2 degrees angular deviation and 45.0 mm distance to target for a 45-mm screw. The dark green circle in the center is used for gaze tracking, for activating virtual buttons (not shown). The three smaller circles are used for high-precision targeting: the dark blue circle is a circle around the bone entry point (green dot); the light blue circle is a circle around the instrument tip. The red circle is located approximately 10 cm up the shaft of the instrument and projected onto the bone entry plane. Because of the distance, it magnifies the angular deviation. The surgeon moves the instrument tip to make the light blue circle (instrument tip) concentric with the dark blue circle (bone entry point); the surgeon then pivots the instrument to make the red circle concentric (magnified angular deviation). The purple box on the left is a graphical representation magnifying the positional deviation from the instrument tip to the green bone entry point 10 times. As the surgeon moves the instrument tip, the intent is to “nail” the green bone entry point with the purple box (10× magnified positional deviation of instrument tip), achieving a positional deviation close to 0 mm. Note that the targeting circles and purple box are directly superimposed onto the surgical site and centered around the instruments; thus, they are in the center of the surgeons’ view while they advance the instruments. An example of the AR display is shown here in the sawbones model used for surgeon training at the beginning of the study. (C) Example of AR display provided by the novel AR system. The surgeon has moved the instrument tip to make the light blue circle (instrument tip) concentric with the dark blue circle (bone entry point); both are superimposed now. The surgeon has pivoted the instrument to make the red circle also concentric (magnified angular deviation) with the dark blue circle (bone entry point), achieving an angular deviation of 0.2 degrees. The surgeon has “nailed” the green bone entry point with the purple box (10× magnified positional deviation of instrument tip), achieving a positional deviation of 0.3 mm. Note that the targeting circles and purple box are directly superimposed onto the surgical site and centered around the instruments; thus, they are in the center of the surgeons’ view while they advance the instruments. The virtual screws displayed on the CT scan on the right and the instrument trajectory have changed in color from yellow (B) to green. Green indicates a positional deviation of less than 1.5 mm and an angular deviation of less than 3.0 degrees. AR, augmented reality; CT, computed tomography; F, fiducial array; HL, headlight; L, loupes; PD, power drill; PI, physical instrument; RA, rotating adapter; SD, screwdriver. The AR headset is customized for each surgeon. Custom lens inserts are provided with the specific prescription of each surgeon (Zeiss Vision Care, Hebron, KY, USA). The AR headset can optionally be used with surgical loupes and a headlight (Figure 1A). The AR headset comprises a 4k resolution organic light-emitting diode (OLED) display and has a weight of approximately 0.7 pounds (320 g) including head mount and attached tracking arrays. The AR display shows sagittal and axial images, including tracked instruments, as well as novel 3D models and stereoscopic targeting tools which magnify the display of any positional deviation and angular deviation approximately 10 times and which are directly overlaid onto the surgical site in the visual field of the surgeon (Figure 1B,1C). The AR system is configured for using both pre-operative CT as well as intra-operative spin. In the present study, only intra-operative spin was used. An imaging phantom is used for registering intra-operative spins. The imaging phantom is automatically recognized by the software and used to register the 3D imaging data in the coordinate system of the spinal reference frame attached to the patient either via pelvic pin or spinal clamp. Surgical technique For open exposure (25 screws), a midline incision was performed and soft tissue was dissected using standard surgical technique. A spinal clamp was attached to a spinous process, e.g., at L4 or L5 or T12 or T11. For MIS (55 screws), a pelvic pin was placed with a mallet in the posterior superior iliac spine (PSIS) region; alternatively, a small incision was placed over a spinous process for placement of the spinal clamp. The spinal reference frame was then attached to the pelvic pin or spinal clamp. The imaging phantom was attached to the spinal reference frame via a mechanical connection, and a 3D spin of the spine was acquired (O-arm, Medtronic, Minneapolis, MN, USA). Following acquisition of the 3D spin, the imaging phantom was removed from the spinal reference frame. Depending on surgeon preference, the following instruments of the different pedicle screw manufacturers were used and tracked during the surgery: awl, burr (Midas Rex, Medtronic, Minneapolis, MN, USA), Jamshidi needle, drill, tap and screwdriver. Image analysis A post-operative spiral CT (Siemens SomatomGo, Siemens Healthineers, Malvern, PA, USA) was obtained following pedicle screw implantation. The position of each screw was assessed by a fellowship-trained musculoskeletal radiologist (J.P.) blinded to surgeon, approach, and planning data and was graded according to the Heary (13) and Gertzbein-Robbins classifications (14). Heary grades I and II and Gertzbein-Robbins grades A and B and were considered accurate placement (15,16). Positional error (PE) and trajectory angle error of the placed pedicle screws relative to the planned position and trajectory were also determined in three dimensions (3D analysis). The vertebrae visualized on the intra-operative cone beam CT were segmented and registered level by level to the postoperative CT scan using Amira3D software (Thermo Fisher Scientific, Waltham, MA, USA). The PE was calculated as the perpendicular distance between the actual screw tip position and the planned trajectory axis (17). The trajectory angle error was calculated as the angle between the actual screw axis and the planned trajectory axis (17). Statistical analysis All data related to 3D measurements for PE (in mm) and angular error (AE, in degrees) were summarized using descriptive statistics, including mean, standard deviation, and standard error of the mean. Exploratory within-study subgroup analyses were performed to evaluate whether PE and AE differed by implanting surgeon, surgical approach (minimally invasive versus open), and spinal region (thoracic versus lumbosacral). Because the sample size was limited and multiple screws were placed within the same cadaver specimens, these subgroup analyses were considered exploratory and interpreted cautiously. To place the observed PE and AE values in the context of the existing literature, we compared the current study results with previously published trajectory-based quantitative accuracy data from navigation, robotic, and AR systems. These cross-study comparisons were intended as contextual benchmarking rather than as head-to-head comparative evidence of superiority. This approach is consistent with the broader principles of evidence synthesis, in which summary-level data from heterogeneous studies may be used to provide interpretive context when direct comparative trials are unavailable, provided that heterogeneity and its effect on inference are explicitly acknowledged. The percentage difference in PE for the novel AR guidance system relative to published systems was calculated as: [(PE published system − PE novel AR system)/(PE published system)] × 100. Similarly, the percentage difference in AE was calculated as: [(AE published system − AE novel AR system)/(AE published system)] × 100. When sufficient summary statistics were available from published studies, exploratory comparisons of mean PE and AE values were performed using Welch’s unequal-variance t-test because sample sizes and variances differed across studies. However, these comparisons were interpreted cautiously because the underlying studies differed with respect to anatomy, operative conditions, measurement methods, and case mix. All statistical tests were two-sided, with a significance level of 5%. Data analyses and summaries were conducted using RStudio (RStudio Team, 2024. Integrated Development for R. RStudio, PBC, Boston, MA, USA). Results Radiographic placement accuracy Of the 80 screws implanted, 76 had a Heary grade 1 and a Gertzbein Robbins grade A, with the screws located inside the pedicle cortices. Four pedicles were encountered where their pedicle diameter was small and no appropriate screw size was available. In these cases, in-out-in technique was planned with an intended small lateral breach, resulting in a planned Heary grade 2 and Gertzbein Robbins grade B, which was considered accurate placement based on both classifications (13-15). No medial or inferior breaches were observed. There were no instances of screw placement violating the spinal canal or neural foramina. Accordingly, the overall screw placement accuracy was 100% based on the accepted thresholds, with Heary grades 1 and 2 and Gertzbein-Robbins grades A and B considered accurate placement (13-16) (Table 1). Table 1Comparative accuracy of different navigation, robotic, and augmented reality systems based on Heary (13), Gertzbein (14), and Gertzbein Robbins (GR) (14) classificationsSystemReferenceAccuracy rate, %Screws placedNo. of significant breaches (Heary >2, Gertzbein >1, Gertzbein Robbins >B)Navigation systems Brainlab NavigationGanguly et al., 2020 (18)96.6592 Medtronic Stealth StationSadrameli et al., 2018 (19)97.31825 Medtronic Stealth StationVan de Kelft et al., 2021 (20)97.53539 Stryker Navigation Nav3iGubian et al., 2022 (21)100.01400 Seaspine 7D SurgicalJakubovic et al., 2018 (22)94.328916Robotic systems Mazor X Robotvan Dijk et al., 2015 (5)98.24879 Globus Excelsius RobotGodzik et al., 2019 (17)96.61164 Globus Excelsius RobotJiang et al., 2020 (23)100.01840 Globus Excelsius RobotVardiman et al., 2020 (24)97.73488Augmented reality systems Augmedics XvisionMolina et al., 2021 (25)99.11131 Brainlab Magic LeapFrisk et al., 2022 (9)93.3453 Novarad VisARFelix et al., 2022 (8)96.01245 OnPoint AR-AICurrent study100.0800Accurate placement defined as Heary grades 1 and 2, Gertzbein grade 0 and 1, or Gertzbein Robbins grade A and B (15,16). Comparative accuracy of different navigation, robotic, and augmented reality systems based on Heary (13), Gertzbein (14), and Gertzbein Robbins (GR) (14) classifications Accurate placement defined as Heary grades 1 and 2, Gertzbein grade 0 and 1, or Gertzbein Robbins grade A and B (15,16). Quantitative error measurements Based on 3D analysis of screw trajectories planned on intra-operative cone beam CT images compared to post-operative CT (Figure 2), PE and AE (mean ± standard deviation) for the novel AR system were 1.1±0.5 mm and 1.3°±0.5°, respectively (Figure 3A,3B). Range for PE was 0.22–2.72 mm and range for AE was 0.39–3.85°. Percent increment in PE and AE for different navigation, robotic, and AR systems compared to the novel AR system ranged between 23.7–80.6% and 40.6–79.5%, respectively, with the differences being statistically highly significant for all comparisons (Figure 3C,3D). Example of 3D reconstruction intraoperative versus post operative. (A) An overlay of a 3D reconstruction of an intra-operative cone beam CT (cyan color) with a post-operative CT (red color). The trajectories (pink) for the pedicle screws were planned by the surgeons based on the intra-operative cone beam CT. (B) The implanted pedicle screws (cyan color) in the post-operative CT data. The trajectories (pink) planned on the intra-operative cone beam CT are also visualized. (C) The implanted pedicle screws (cyan color) in the post-operative CT data with greater transparency. The implanted pedicle screws have been placed aligned with and superimposed onto the planned trajectories (pink). 3D, three-dimensional; CT, computed tomography. PE published systems compared to novel AR-AI system. (A) The PE in mm and 1 standard deviation for different navigation, robotic, and augmented reality systems based on published studies (5,8,9,11,17,18,21,23,26-28). One study (11) did not publish SD, but confidence intervals instead (not shown). (B) The AE in degrees and 1 standard deviation for different navigation, robotic, and augmented reality systems based on published studies (5,8,9,11,17,18,21,23,26-28). One study (11) did not publish SD, but confidence intervals instead (not shown). (C) The percent increment in PE of existing navigation, robotic and AR systems compared to the novel AR system and the statistical significance level for the different comparisons. (D) The percent increment in AE of existing navigation, robotic and AR systems compared to the novel AR system and the statistical significance level for the different comparisons. AE, angular error; AI, artificial intelligence; AR, augmented reality; PE, positional error; SD, standard deviation. Differences between surgeons, open vs. MIS surgery, thoracic vs. lumbar spine Exploratory subgroup analyses demonstrated no statistically significant differences in positional or AE between surgeons (PE: P=0.29; AE: P=0.43), indicating consistency of performance across operators. However, statistically significant differences were observed between surgical approaches, with differences in both PE (P=0.02) and AE (P=0.02), with MIS being slightly more accurate than open surgery, as well as between thoracic and lumbosacral screw placement (PE: P=0.003; AE: P=0.003). These findings should be interpreted cautiously given the exploratory nature of the analysis and the limited sample size. Discussion A variety of imaging and navigation technologies have been developed to help improve the accuracy of pedicle screw placement and related complications (4-9,11,17,18,21-23,26,27). Screw malposition can be significantly reduced from 10.3% with freehand technique using intra-operative fluoroscopy only to 3.6% with navigation using intra-operative CT imaging (P<0.001) (3). In addition, recent studies comparing navigation using intra-operative cone beam CT with freehand technique using fluoroscopic guidance indicate a decreased risk of reoperation for hardware failure (2.9% vs. 5.9%, P<0.01), screw malposition (1.6% vs. 4.2%, P<0.01), and all-cause reoperation (5.2% vs. 10.9%, P<0.01) (29). Traditional navigation with intra-operative CT or cone beam CT imaging is, however, hampered by high cost, which can amount to several hundred thousand dollars (4), in addition to personnel costs in the OR, which has constrained the distribution of these technologies. Capital equipment expenditure is of even greater concern with the introduction of robots for pedicle screw placement (5,6,17), which can cost more than $1 million in capital in addition to service contracts, maintenance contracts and disposables (4). A meta-analysis of five different studies comparing robot-assisted and conventional freehand pedicle screw placement in a total of 257 patients and 1,105 screws found that there was no statistical difference in the accuracy between robot-assisted and conventional freehand pedicle screw placement regardless of using 0 (screw within pedicle) or 2 mm (breach of <2 mm) grading criteria. Robotic pedicle screw placement is also not free of technical challenges, which can require supplemental fluoroscopy (30) and can still result in imperfect screw positioning (5,17,23,31). The reported clinical accuracy of most navigation, robotic, and AR systems is high, with a radiographic accuracy rate of >95% reported for most systems (Table 1). However, this means that even with significant breaches only in the 2–5% range with most of these systems (Table 1), on an annual spinal fusion procedure volume of 1.36 million in the U.S. which includes approximately 900,000 pedicle screw procedures (1), between 18,000 to 45,000 patients would be at risk to experience complications related to screw malposition including pain, nerve root irritation, or motor weakness (2); the actual number is potentially higher since each of these procedures require implantation of 4 or more pedicle screws. Accurate screw placement is of critical importance in difficult anatomic situations such as in cervical and thoracic spine procedures, revision spine cases and complex deformities. The novel AR guidance system had a radiographic, CT-based screw placement accuracy of 100%, without any medial or inferior breaches and no instances of screw placement violating the spinal canal or neural foramina. PE and AE were remarkably small with the novel AR guidance system, with values of 1.1±0.5 mm and 1.3°±0.5°, respectively (Figure 3A,3B). This means that in absolute numbers the novel AR guidance system was between 2 and 5 times more accurate than data reported for existing navigation (18,21,26-28), robotic (5,17,23), and AR systems (8,9,11), with the differences being statistically highly significant for all comparisons (Figure 3C,3D). When applying a Bonferroni adjustment for the 19 comparisons (in 3 of the comparison studies, either PE or AE were not reported, resulting in a total of 19 comparisons), the adjusted significance level changes from P<0.05 to P<0.0026. Even with this adjusted significance level, all but 2 of the comparisons remain statistically significant. Notably, the Bonferroni adjustment can be overly conservative when the number of tests is large, as in this study, leading to an increased risk of type II errors (false negatives) (32). While these results are based on cadaveric data only, potential reasons why the novel AR guidance system offers such a high level of procedural accuracy appear to be multifactorial. Tactile sense Unlike with robotic systems, surgeons retain their tactile sense while positioning or advancing instruments or implants when using AR. While robotic-assisted screw placement has been reported to have high accuracy, skiving has been pointed out as a primary factor for deviations from the surgical plan (33). Irregular bone surfaces, steep entry point angles, soft tissue pressure, for example in obese patients, high drilling pressure, and dull drill bits have been reported to induce skiving (34-37). With AR, the surgeons retain their tactile feel and they will readily perceive the skiving through their tactile senses and then use the visual AR display to correct any resultant PE and AE. The novel AR system is configured to visualize any misalignment of an instrument, screw or other implant for the surgeon with regard to both PE and AE, thereby facilitating any correction or adjustment. Simultaneous direct viewing of surgical site and AR information Direct viewing of the surgical site paired with the AR display and guidance of the surgical instruments may contribute to a high placement accuracy of pedicle screws. With navigation or robotics, imaging studies of the patient are displayed in the OR on an external computer monitor, often several feet away from the patient and surgeon (3,5,7,17). Since the surgical field is in a different location and has different coordinates than the computer monitor, hand-eye coordination can be challenging for the surgeon and requires a significant learning period. AR systems are ideally suited to improve hand-eye coordination by overlaying real-time trajectories onto the surgical field (8,9,11). The direct viewing of the surgical site through the AR display with simultaneous display of the navigation information resolves issues related to fatigue from looking back and forth between the patient and the computer monitor and inadvertent hand movements while viewing the computer screen. Unlike existing AR systems (8,9,11), the novel AR guidance system used in this study provides 3D stereoscopic displays that are configured not only to display but to magnify any positional deviation and angular deviation approximately 10 times (Figure 1B,1C). These 3D stereoscopic representations are directly overlaid onto the surgical site and are not only visible to the surgeons while they advance the instruments but centered around the instruments (Figure 1B,1C); thus, the instruments can be finely adjusted by the surgeons with use of the 10× magnified display of positional deviation and angular deviation while the instruments or implants are being advanced. Ergonomics Surgeons frequently wear intra-operative loupes (38), in spinal surgery, for example, for decompressions or fine bone resection. Currently available AR headsets are not compatible with the use of intra-operative surgical loupes (8,9,11,38); the headset has to be taken off before the surgeon can place the loupes on their head. The novel AR headset integrates surgical loupes and a headlight with the AR display (Figure 1A). Similar to the AR display, the surgical loupes are customized for the surgeon. By integrating optical loupes with the AR display system, the surgeon can readily move from “real world” magnified view of the spine through the loupes, a view surgeons use every day today in their clinical practice, to the augmented view, simply by moving their eye down to the loupe or up to the AR display, where the AR display provides a magnified 3D stereoscopic representation of positional deviation and angular deviation of the surgical instruments, which is synergistic with the magnified loupe view of the physical spine, which may have contributed to the high level of surgical accuracy observed in this study. Display resolution, accuracy of AR overlay Low-resolution AR and VR displays can lead to poor image quality and ineffective blurred or pixelated virtual displays (39). Currently available AR systems for spinal surgery have a display resolution of 1,280×720 pixels per eye (11) or 1,440×936 pixels per eye (8,40,41). A blurred or pixelated appearance of CT scans by a lower resolution AR display may limit the visual acuity of the display and may adversely affect procedural accuracy. The AR headset used in this study has a display resolution of 1,920×1,080 pixels per eye utilizing high resolution micro-OLEDs. One of the most important limitations of AR displays by see-through optical head-mounted displays is the lack of accuracy when the AR screen overlays the virtual models on the physical world (10). With existing AR headsets (8,9,11), every time the surgeon places the headset on their head at the beginning of the procedure, the position of the headset and with that the position of the left and right eye AR display units and the virtual displays will shift or move and be in a different position in relationship to the surgeon’s eyes, thus directly affecting the accuracy of the overlay or superimposition of the virtual display with the spine. The novel AR system used in this study is different in this regard since it comprises an electronic display alignment, which is configured to align virtual objects with the corresponding physical objects, thereby ensuring an accurate overlay irrespective of the position and orientation with which the surgeon places the headset on their head (Figure 1B,1C). AR headset weight Existing AR headsets weigh between 1.3 pounds (566 g) (40) and 3 pounds (1,360 g) (11). A high weight of the AR headset worn by the surgeon can lead to fatigue and surgeon discomfort (42), which may adversely affect surgical accuracy, in particular during long procedures. The AR headset used in the current study weighs 0.7 pounds (320 g), which includes the head mount and attached tracking arrays. Fatigue or discomfort were not reported by the five surgeons wearing the headsets during this study. Scanner tracking vs. image phantom for image registration Modern navigation systems (18,27,28) typically track the intra-operative CT or cone beam CT scanner by detecting reflective markers or light-emitting diodes (LEDs) placed on the gantry of the scanner during the scan (43). The resultant scanner tracking data are then used to register the scan in the coordinate system of the spinal reference frame array. The accuracy of this registration approach is dependent on the positional accuracy of the tracking system, which can vary between 0.22 and 1.02 mm for different tracking systems (44); other factors such as the distance from the tracking camera to the scanner may also affect the accuracy. The AR system used in this study does not utilize scanner tracking, but is registering the intra-operative CT or cone beam CT scan using an imaging phantom that is mechanically connected to the spinal reference frame during the scan. The tolerance of the mechanical connection is approximately 0.02 mm and 0.15 degrees. While this study did not provide a direct side-by-side comparison of optical scanner tracking vs. the use of an image phantom attached to the spinal reference frame array, the tight tolerance of the mechanical connection may be an additional factor contributing to the observed accuracy in placing pedicle screws. Limitations This study has several limitations. First, it was conducted in cadaveric specimens and therefore reflects a more controlled environment than live clinical surgery. Although cadaver studies are appropriate for technical validation, the present findings should not be interpreted as direct evidence of clinical benefit, safety, workflow improvement, or cost-effectiveness in patients. Second, the comparisons with previously published navigation, robotic, and AR studies are limited by substantial heterogeneity across studies, including differences in specimen or patient characteristics, vertebral levels instrumented, operative techniques, surgeons, screw systems, imaging protocols, and methods used to quantify accuracy. For this reason, the present cross-study analyses should be interpreted as exploratory contextual benchmarks rather than as definitive comparative evidence. Our intent was not to perform a formal meta-analysis or indirect treatment comparison, but rather to benchmark the magnitude of the observed PE and AE values against previously published trajectory-based studies. Third, some of the published studies used less comprehensive measurement approaches than the present investigation. For example, Miller et al. (28), Oertel et al. (27), and van Dijk et al. (5) measured PE and AE on axial or sagittal postoperative imaging only, whereas the present study used true 3D analysis. Such methodological differences may affect the absolute magnitude of reported errors and further limit direct comparability. Fourth, the within-study subgroup analyses by surgeon, surgical approach, and spinal region were exploratory and should be interpreted cautiously because of the limited sample size and the clustering of multiple screws within the same cadaver specimens. These analyses were included to provide internal context, but the study was not powered primarily for subgroup inference. Moreover, the present study evaluated only thoracic and lumbosacral pedicle screw placement and did not evaluate other spinal indications, operative workflows, ergonomic endpoints, or cost efficiency. Future studies should assess the performance of the system in larger clinical cohorts, include independent validation, and evaluate workflow, usability, and cost-related outcomes directly. Finally, this report did not assess accuracy over time (first vs. last screws), which needs to be assessed in follow-on studies. M.A., H.U.B., and S.S.H. served on the advisory board of the company that developed the 3D stereoscopic AR system, although they had not received compensation at the time the data were acquired. Conclusions In conclusion, the novel AR guidance system with 3D stereoscopic targeting demonstrated high accuracy in pedicle screw placement when compared to published studies on existing navigation, robotic, and AR technologies. The ability to achieve a PE of 1.1 mm and an AE of 1.3° makes this technology particularly suited for small or narrow pedicles and challenging anatomy. Supplementary The article’s supplementary files as Available at https://jss.amegroups.com/article/view/10.21037/jss-2026-1-0034/prf