conventional pedicle screw placement has been performed via the freehand technique using anatomical landmarks to identify the entry point and trajectory. Although freehand insertion accuracy rates have been reported to be as high as 98.3%,1 misplaced screws can cause significant morbidity, with neurological, dural, vascular, and visceral injuries.2–4 Recent advances in imaging and navigation to minimize screw misplacement include fluoroscopy-guided, CT-guided, and robot-assisted techniques.5,6 However, limitations of these technologies include line-of-sight interruption and attention shift.7,8
Augmented reality (AR) describes the use of a computer-generated image displayed on the user’s view of the real world, possibly mitigating the limitations of other navigational systems. AR is a novel technology that has emerged for neuronavigation in orthopedic surgery, hepatobiliary surgery, and urogenital surgery.9,10 When applied to spine surgery, it offers the potential for efficient, safe, and accurate placement of spinal instrumentation.11,12 Previous studies have explored the accuracy of existing navigation and robotic technologies for the placement of pedicle screws in humans.6,13 We present the first single-institution study on the accuracy of 205 consecutive thoracic, lumbar, and sacral pedicle screws placed in patients by using an FDA-approved AR head-mounted display (HMD) technology.
Methods
Data Collection
After obtaining institutional review board approval, we retrospectively reviewed data obtained in all patients who underwent consecutive AR-assisted pedicle screw placement performed at a single academic tertiary care center between June 1, 2020, and December 1, 2020. Inclusion criteria included age older than 18 years and placement of pedicle screws in the thoracic, lumbar, and/or sacral spine. S2-alar-iliac (S2AI) screws or other iliac fixation screws were excluded from this study. All patients were treated by 3 senior authors (S.L.L., D.M.S., and T.F.W.). Demographic factors included sex, age, and ethnicity. Clinical information included prior spine surgery, preoperative symptoms, and imaging findings. Instrumentation information included level and number of screws placed.
Surgical Technique and Operative Workflow
Patients were positioned prone on a Jackson table with intraoperative monitoring and prepped and draped in standard sterile fashion. All cases in this series were performed with open exposure. After exposure, a registration clamp and marker were placed on a spinous process—the level of which was chosen based on the exposure and number of planned instrumentation levels. Intraoperative 3D imaging and registration was acquired with an O-arm (Medtronic) or C-arm (Ziehm) unit. The registration marker on the clamp was exchanged for a tracking marker that was placed on the contralateral working side. Surgeons were fitted with an AR-HMD (xvision; Augmedics), and pedicle screws were placed with AR computer navigation assistance (Figs. 1 and 2). The initial entry point was selected using anatomical landmarks and confirmed with the AR computer navigation. Cannulation of the pedicles was completed with a tracked pedicle probe and pedicle tap, or with a tracked power tap alone. Pedicle screws were also then placed using AR-HMD assistance manually or with a power driver. The registration clamp and patient marker were removed. The rest of the surgery (which may have included osteotomies, discectomy, interbody placement, corpectomy, and/or tumor resection) proceeded in standard fashion. Confirmatory postplacement CT was obtained either intraoperatively with an O-arm or postoperatively with conventional CT.
Left: Surgeons (A.L. and T.F.W.) are fitted with an AR-HMD with direct visualization of the surgical field. Right: The tracking marker (double arrows) is clamped onto a spinous process. The instrument is navigated with placement of a reflective marker (single arrow). Copyright Augmedics. Published with permission. Figure is available in color online only.
A live intraoperative view through the HMD. The navigational data are overlaid directly on the operative field. Copyright Augmedics. Published with permission. Figure is available in color online only.
Screw Accuracy Assessment
Screw accuracy was assessed using the Gertzbein-Robbins scale and defined as follows: grade A, perfect intrapedicular localization without any cortical breach; grade B, < 2-mm pedicle breach; grade C, < 4-mm pedicle breach; grade D, < 6-mm pedicle breach; or grade E, ≥ 6-mm pedicle breach (Fig. 3).14 Accuracy was assessed using the patient’s confirmatory CT scan by a neuroradiologist (M.K.) who was blinded to the patient’s identity, medical history, and technique of pedicle screw placement. Grades A and B were designated as accurately placed.
Examples illustrating the Gertzbein-Robbins grading scale. A: The right-hand screw is rated grade A and the left-hand screw is rated grade B. B: The left-hand screw is rated grade C. C: The left-hand screw is rated grade D.
Results
Baseline Characteristics
A total of 32 patients were intended to undergo AR-assisted pedicle screw placement; however, the use of AR was aborted in 3 cases. One patient’s registration imaging could not be transferred to the system’s console. In 2 patients, the registration clamp was too short due to body habitus. One patient underwent S2AI screw placement alone with AR assistance. These 4 patients were excluded from our analysis.
Twenty-eight consecutive patients (with the exception of the above exclusions) underwent thoracic, lumbar, or sacral pedicle screw placement with AR assistance (Table 1). Eleven (39%) patients were male and 17 (61%) were female. The majority of patients were Caucasian (n = 27, 96%). The median age at time of surgery was 62.5 (IQR 13.8, range 38–84) years. The median body mass index was 31 (IQR 8.6) kg/m2. Fifteen (54%) patients had prior spine surgeries—4 (27%) of whom had decompression surgeries alone, whereas 11 (73%) patients had prior fusion surgeries.
Demographics and clinical information
| Variable | Value |
|---|---|
| Total pts | 28 |
| Female, no. (%) | 17 (61%) |
| Caucasian, no. (%) | 27 (96%) |
| Age, yrs (IQR) | 62.5 (13.8) |
| BMI, kg/m2 (IQR) | 31 (8.6) |
| Prior spine surgery, no. (%) | 15 (54%) |
| Decompression alone | 4 (27%) |
| Fusion | 11 (73%) |
| Indication for surgery, no. (%) | |
| Degenerative | 12 (43%) |
| Deformity | 12 (43%) |
| Tumor | 3 (11%) |
| Trauma | 1 (4%) |
| Presenting symptoms, no. (%) | |
| Low-back pain | 26 (93%) |
| Radicular leg pain | 19 (68%) |
| Weakness | 10 (36%) |
BMI = body mass index; pts = patients.
Indications for surgery included degenerative disease (n = 12, 43%); deformity correction (n = 12, 43%); tumor (n = 3, 11%); and trauma (n = 1, 4%). The majority of patients (n = 26, 93%) presented with low-back pain and 19 (68%) patients presented with radicular leg pain. Ten (36%) patients had documented lower extremity weakness on preoperative examination. In 5 cases, the weakness was mild and believed to be potentially pain related.
Screw Placement Accuracy
A total of 205 screws were placed, with a median of 7 (IQR 4, range 2–15) screws placed per patient. Approximately half of the screws (n = 112, 55%) were placed in the lumbar spine, 67 (33%) screws were placed in the thoracic spine, and 26 (13%) screws were placed at S1 (Table 2).
Screw accuracy
| Variable | Value |
|---|---|
| Total screws, no. (%) | 205 |
| Lumbar | 112 (55%) |
| Thoracic | 67 (33%) |
| S1 | 26 (13%) |
| Screws per pt (IQR) | 7 (4) |
| Overall accuracy | 98.0% |
| Thoracic | 98.5% |
| Lumbar/S1 | 97.8% |
| GRS grade, no. (%) | |
| A | 193 (94%) |
| B | 8 (4%) |
| C | 3 (1.5%) |
| D | 1 (0.5%) |
| E | 0 (0%) |
GRS = Gertzbein-Robbins scale.
Screw placement accuracy was 98.5% for thoracic screws, 97.8% for lumbar/S1 screws, and 98.0% overall. Of the 205 screws placed, 193 (94%) were grade A, 8 (4%) were grade B, 3 (1.5%) were grade C, 1 (0.5%) was grade D, and no screws were grade E. Patients with screws other than grade A are listed in Table 3. All grade C and D screws were clinically asymptomatic.
Screw placement accuracy for patients with grades B–D based on the GRS
| Level | Pt 4 | Pt 8 | Pt 13 | Pt 17 | Pt 25 | Pt 28 | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Lt | Rt | Lt | Rt | Lt | Rt | Lt | Rt | Lt | Rt | Lt | Rt | |
| T1 | A | A | ||||||||||
| T2 | A | A | ||||||||||
| T3 | ||||||||||||
| T4 | B | A | B | |||||||||
| T5 | B | C* | A | A | ||||||||
| T6 | B | A | A | |||||||||
| T7 | A | A | A | |||||||||
| T8 | A | |||||||||||
| T9 | A | |||||||||||
| T10 | A | |||||||||||
| T11 | A | |||||||||||
| T12 | B | |||||||||||
| TSA subtotal | 91% | 100% | ||||||||||
| L1 | A | A | ||||||||||
| L2 | A | A | ||||||||||
| L3 | A | A | B | A | B | A | ||||||
| L4 | † | A | C* | A | A | B | D* | A | ||||
| L5 | † | C* | A | A | A | A | A | A | ||||
| S1 | A | A | A | A | ||||||||
| LSA subtotal | 100% | 75% | 88% | 100% | 83% | |||||||
LSA = lumbar screw accuracy; TSA = thoracic screw accuracy.
Grades C and D are italicized to visually emphasize the few screws that were categorized as unsatisfactory according to the GRS.
Of note, in the patient in case 13 the left L4 and L5 screws were placed via a freehand technique.
Discussion
Accuracy of Pedicle Screw Placement
In the present study we found an accuracy rate of 98.0% for AR-assisted pedicle screw placement in which unique HMD technology was used. These first 205 screws were consecutively placed. Similar studies have been published on the accuracy of other robotic or navigational systems. In 2015, van Dijk et al. described minimally invasive screw placement in 112 patients in whom they used the SpineAssist robotic platform developed by Mazor, which demonstrated a 97.9% accuracy (Gertzbein-Robbins grades A and B).15 The accuracy was improved to 98.5% with the introduction of the Renaissance system,16 and a reported 98.7% (Gertzbein-Robbins grade A) with the subsequent Mazor X release.17 ExcelsiusGPS robotic platform accuracy has been similarly well documented in the literature, with Huntsman et al., Godzik et al., and Jiang et al. reporting 99%, 96.6%, and 100% pedicle screw accuracy (Gertzbein-Robbins grades A and B), respectively.18–20 The ROSA Spine robotic assistant has had more limited accuracy data but a prospective, case-matched study comparing ROSA with freehand screw placement found ROSA screws to be 97% accurate.21 Competitor TIANJI has also been reported to have up to 98.7% accuracy in prospective randomized controlled studies.22,23 A recent systematic review including 78 studies with 7858 patients examining the accuracy for freehand, fluoroscopy-assisted, CT navigation–guided, and robotic-assisted techniques reported rates of 93.1%, 91.5%, 95.5%, and 90.5%, respectively.6 In this early report, our accuracy with AR-HMD is 98.0%. Although our accuracy rate compares favorably to other navigation and robotic modalities for pedicle screw placement, further studies are required to directly compare the accuracy of this new modality to existing navigation modalities.
Limitations of the Technology
This technology is not immune to the causes of failure with other navigation systems including failure of registration hardware/software, failure to obtain adequate fluoroscopic images,24 and disruption of the interspinous clamp used for registration. In this first cohort of patients, we have experienced several of these limitations. First, given the increased worldwide prevalence of obesity,25 we have encountered two cases in which the use of AR assistance was aborted due to large body habitus, because the registration clamp was too short to extend above the soft tissue of the patient’s back. Second, unlike other systems that can use a preoperative CT for registration, the current version of this technology requires an intraoperative CT scan for registration and is thus limited by the ability to obtain this scan in the operating room. Due to one patient’s large body habitus, the initial intraoperative CT was too low quality to create a plan with the software, and a second intraoperative CT had to be obtained. In the third instance in which the use of AR was aborted, the intraoperative scan was unable to be transferred to the console and thus instrumentation placement was converted to a freehand technique. Finally, the system relies on an interspinous clamp that lies in the operative field and can be inadvertently disrupted. Improvements to the system (including the development of registration markers to be placed in more rigid locations such as the posterior superior iliac spine, and the use of preoperative CT for registration) are currently being developed. Longer spinous process clamps have already been developed, approved, and used in patients.
One unique limitation of the Augmedics xvision AR-HMD system is associated with the fact that any instrument can be universally navigated with the placement of reflective markers. In one of our patients we encountered an issue in which the reflective marker slid down the screwdriver during screw placement, causing the navigation to become inaccurate. Fortunately, because the pedicles had already been cannulated, the screws were easily placed into the pedicle tracts.
Other adoption limitations include mechanical discomfort with wearing the headset, visual discomfort, and visual obstruction of the surgical field. With initial use of the AR-HMD, users may experience sensory overload and must adapt to the direct projection of images on the retina mixed with real visual objects. Because the headset prevents the concurrent use of a surgical headlight, visualizing the surgical field and key anatomical structures can be difficult. Early on in our series, we encountered this issue when attempting to drill pilot holes while wearing the AR-HMD gear. To address this, in our current workflow the location of the pilot holes is decided upon and drilled based on anatomical landmarks, with the use of a surgical headlight prior to obtaining the intraoperative registration CT. The location of the pilot holes is then verified with the AR-HMD and can be adjusted as needed. One other alternative is the use of a navigated drill for pilot hole placement. Modifications of the current headset with the addition of magnification and a headlight are already being developed.
Benefits and Future Vision of the Technology
Despite these limitations, we believe that the benefits of AR-HMD are significant and allow for improvements upon existing navigational systems. One of the biggest advantages of the AR-HMD system is its ability to allow the surgeon to simultaneously visualize the operative field (Fig. 2), rather than looking up at a computer screen. This minimizes attention shift, which has been shown to negatively impact both cognitive and motor tasks.7,8 Additionally, the AR-HMD system minimizes line-of-sight interruption, which is the disturbance of live computer navigation by an obstacle blocking the tracking markers or the remote tracking camera. Because the HMD has an integrated tracking camera, there is a much shorter distance between the camera and tool markers, which limits the possibility of line-of-sight interruption. Another important benefit is that this system is implant and vendor “agnostic,” and almost any instrument can be universally navigated with the placement of reflective markers. Thus, surgeons do not have to change their preferred type of instrumentation to use this system (in contrast, the use of the ExcelsiusGPS robot requires use of Globus instrumentation). Finally, the pricing of this system is not cost-prohibitive. At the time of this report, the cost of the Augmedics xvision system (including two headsets, the computer, and instrument sets) is approximately $150,000, with disposables costing approximately $1500 per case.26
The current AR-HMD system is in its nascent stages, and we anticipate that many of the technical limitations that we have described above will quickly improve with future iterations of the technology. As engineers continue to improve upon the headset itself, issues such as comfort, a built-in light source, and built-in lens magnification will no longer be limitations to this technology’s use. Additionally, registration with preoperative CT scans will be likely to replace the need for an intraoperative CT scan.
Based on our early experience, the advantages of the AR-HMD outweigh its limitations and allow for safe placement of pedicle screws. As this new navigational system becomes more widely adopted, we envision a multitude of uses for it beyond just pedicle screw placement. Already, we are exploring the use of AR for other indications, including the placement of S2AI screws and planning of spinal tumor resections and osteotomies.27 We also hope that this technology can be merged with other existing robotic systems.
Study Limitations and Future Directions
The major limitation of this study lies in the fact that our data represent only the first 205 pedicle screws at a single institution under the guidance of 3 experienced attending physicians, which we expect will be rectified as more cases are performed and AR is more widely adopted for spinal instrumentation placement. The majority of our cases were performed for degenerative or deformity pathologies, with a small fraction of patients undergoing surgery due to tumor or trauma. Furthermore, all cases were performed by attending physicians who have had extensive experience with freehand pedicle screw placement as well as other modalities of spinal navigation, which does not allow for generalization across surgeons with varying levels of experience. All cases in this study were performed using an open technique with full exposure of the relevant anatomy, and therefore accuracy of AR-HMD in percutaneous or minimally invasive spine modalities could not be assessed.
Future directions for this new technology include continued evaluation of accuracy, as well as precision, which has emerged as an objective metric for determining if an implant is positioned precisely where intended. The present system’s precision has been previously analyzed in a cadaveric setting of percutaneously placed screws, with a mean screw tip linear deviation of 1.98 mm (99% CI 1.74–2.22 mm) and a mean angular error of 1.29° (99% CI 1.11°–1.46°) from the projected trajectory.12 In a report of the first in-human use of AR (which is included in this series), technical precision analysis determined a mean linear deviation of 2.07 mm (95% CI 1.62–2.52 mm) and angular deviation of 2.41° (95% CI 1.57°–3.25°).28 Further studies of the learning curve and the utility of AR-HMD for resident training are also needed.
Conclusions
AR-assisted spinal instrumentation placement with a unique HMD navigation system is a novel and clinically accurate technology for the insertion of pedicle screws in spine surgery. We describe the first 205 AR-assisted thoracic, lumbar, and sacral pedicle screws consecutively placed at our institution with an accuracy of 98.0% as determined by a Gertzbein-Robbins grade of A or B.
Disclosures
Dr. Molina is a consultant for Augmedics and DePuy. Dr. Sciubba is a consultant for Augmedics, DePuy-Synthes, Stryker, and Baxter. Dr. Witham is a consultant for, investor in, and medical advisory board member for Augmedics. Augmedics did not provide any financial support for the performance of these cases. In addition, they did not have any input into the writing, data interpretation, editing, or submission of this manuscript.
Author Contributions
Conception and design: Witham, Liu, Jin, Westbroek, Ehresman, Pennington, Lo, Sciubba, Molina. Acquisition of data: Witham, Liu, Jin, Cottrill, Lo, Sciubba. Analysis and interpretation of data: Witham, Liu, Jin, Cottrill, Khan, Lo, Sciubba. Drafting the article: Witham, Liu, Jin, Lo, Sciubba. Critically revising the article: Witham, Liu, Jin, Khan, Westbroek, Lo, Sciubba, Molina. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: Witham. Statistical analysis: Witham, Liu. Administrative/technical/material support: Witham, Liu, Jin. Study supervision: Witham, Liu, Jin, Lo, Sciubba.
References
- 1↑
Parker SL, McGirt MJ, Farber SH, Amin AG, Rick AM, et al. Accuracy of free-hand pedicle screws in the thoracic and lumbar spine: analysis of 6816 consecutive screws. Neurosurgery. 2011;68(1):170–178.
- 2
Gautschi OP, Schatlo B, Schaller K, Tessitore E. Clinically relevant complications related to pedicle screw placement in thoracolumbar surgery and their management: a literature review of 35,630 pedicle screws. Neurosurg Focus. 2011;31(4):E8.
- 3
Kayacı S, Cakir T, Dolgun M, Cakir E, Bozok Ş, et al. Aortic injury by thoracic pedicle screw. When is aortic repair required? Literature review and three new cases. World Neurosurg. 2019;128:216–224.
- 4
Parker SL, Amin AG, Santiago-Dieppa D, Liauw JA, Bydon A, et al. Incidence and clinical significance of vascular encroachment resulting from freehand placement of pedicle screws in the thoracic and lumbar spine: analysis of 6816 consecutive screws. Spine (Phila Pa 1976).2014;39(8):683–687.
- 5↑
Kochanski RB, Lombardi JM, Laratta JL, Lehman RA, O’Toole JE. Image-guided navigation and robotics in spine surgery. Neurosurgery. 2019;84(6):1179–1189.
- 6↑
Perdomo-Pantoja A, Ishida W, Zygourakis C, Holmes C, Iyer RR, et al. Accuracy of current techniques for placement of pedicle screws in the spine: a comprehensive systematic review and meta-analysis of 51,161 screws. World Neurosurg. 2019;126:664–678.e3.
- 7↑
Léger É, Drouin S, Collins DL, Popa T, Kersten-Oertel M. Quantifying attention shifts in augmented reality image-guided neurosurgery. Healthc Technol Lett. 2017;4(5):188–192.
- 8↑
Rahmathulla G, Nottmeier EW, Pirris SM, Deen HG, Pichelmann MA. Intraoperative image-guided spinal navigation: technical pitfalls and their avoidance. Neurosurg Focus. 2014;36(3):E3.
- 9↑
Fida B, Cutolo F, di Franco G, Ferrari M, Ferrari V. Augmented reality in open surgery. Updates Surg. 2018;70(3):389–400.
- 10↑
Vávra P, Roman J, Zonča P, Ihnát P, Němec M, et al. Recent development of augmented reality in surgery: a review. J Healthc Eng. 2017;2017:4574172.
- 11↑
Molina CA, Theodore N, Ahmed AK, Westbroek EM, Mirovsky Y, et al. Augmented reality-assisted pedicle screw insertion: a cadaveric proof-of-concept study. J Neurosurg Spine. 2019;31(1):139–146.
- 12↑
Molina CA, Phillips FM, Colman MW, Ray WZ, Khan M, et al. A cadaveric precision and accuracy analysis of augmented reality-mediated percutaneous pedicle implant insertion. J Neurosurg Spine. 2020;34(2):316–324.
- 13↑
Vadalà G, De Salvatore S, Ambrosio L, Russo F, Papalia R, Denaro V. Robotic spine surgery and augmented reality systems: a state of the art. Neurospine. 2020;17(1):88–100.
- 14↑
Gertzbein SD, Robbins SE. Accuracy of pedicular screw placement in vivo. Spine (PhilaPa 1976).1990;15(1):11–14.
- 15↑
van Dijk JD, van den Ende RP, Stramigioli S, Köchling M, Höss N. Clinical pedicle screw accuracy and deviation from planning in robot-guided spine surgery: robot-guided pedicle screw accuracy. Spine (Phila Pa 1976).2015;40(17):E986–E991.
- 16↑
Onen MR, Simsek M, Naderi S. Robotic spine surgery: a preliminary report. Turk Neurosurg. 2014;24(4):512–518.
- 17↑
Khan A, Meyers JE, Siasios I, Pollina J. Next-generation robotic spine surgery: first report on feasibility, safety, and learning curve. Oper Neurosurg (Hagerstown). 2019;17(1):61–69.
- 18
Huntsman KT, Ahrendtsen LA, Riggleman JR, Ledonio CG. Robotic-assisted navigated minimally invasive pedicle screw placement in the first 100 cases at a single institution. J Robot Surg. 2020;14(1):199–203.
- 19
Godzik J, Walker CT, Hartman C, de Andrada B, Morgan CD, et al. A quantitative assessment of the accuracy and reliability of robotically guided percutaneous pedicle screw placement: technique and application accuracy. Oper Neurosurg (Hagerstown). 2019;17(4):389–395.
- 20
Jiang B, Pennington Z, Zhu A, Matsoukas S, Ahmed AK, et al. Three-dimensional assessment of robot-assisted pedicle screw placement accuracy and instrumentation reliability based on a preplanned trajectory. J Neurosurg Spine. 2020;33(4):519–528.
- 21↑
Lonjon N, Chan-Seng E, Costalat V, Bonnafoux B, Vassal M, Boetto J. Robot-assisted spine surgery: feasibility study through a prospective case-matched analysis. Eur Spine J. 2016;25(3):947–955.
- 22↑
Fan M, Lui Y, Tian W. Internal fixation in upper cervical spinal surgery: a randomized controlled study. In: Tian W, Rodriguez y Baena F, eds.The 18th Annual Meeting of the International Society for Computer Assisted Orthopaedic Surgery; June 6–9, 2018;Beijing, China.CAOS; 2018:51-55.
- 23↑
Han X, Tian W, Liu Y, Liu B, He D, et al. Safety and accuracy of robot-assisted versus fluoroscopy-assisted pedicle screw insertion in thoracolumbar spinal surgery: a prospective randomized controlled trial. J Neurosurg Spine. 2019;30(5):615–622.
- 24↑
Joseph JR, Smith BW, Liu X, Park P. Current applications of robotics in spine surgery: a systematic review of the literature. Neurosurg Focus. 2017;42(5):E2.
- 25↑
WHO. Obesity and overweight. Accessed July 13, 2021.http://www.who.int/mediacentre/factsheets/fs311/en/
- 26↑
Dibble CF, Molina CA. Device profile of the XVision-spine (XVS) augmented-reality surgical navigation system: overview of its safety and efficacy. Expert Rev Med Devices. 2021;18(1):1–8.
- 27↑
Molina CA, Dibble CF, Lo SL, Witham T, Sciubba DM. Augmented reality-mediated stereotactic navigation for execution of en bloc lumbar spondylectomy osteotomies. J Neurosurg Spine. 2021;34(5):700–705.
- 28↑
Molina CA, Sciubba DM, Greenberg JK, Khan M, Witham T. Clinical accuracy, technical precision, and workflow of the first in human use of an augmented-reality head-mounted display stereotactic navigation system for spine surgery. Oper Neurosurg (Hagerstown). 2021;20(3):300–309.
