Article Text

Download PDFPDF
Purpose-built, head-mounted 3D display for ophthalmic microsurgery: surgical skill performance and evaluation: a pilot study
  1. Edward Korot1,2,3,
  2. Matthew Mark Rolain4,5,
  3. Allan Evans6,
  4. Aristomenis Thanos7,
  5. Christos Bergeles8,
  6. Edward H Wood3,
  7. Mark A Rolain1
  1. 1Beaumont Eye Institute, Beaumont Health System, Royal Oak, Michigan, USA
  2. 2Moorfields Eye Hospital, London, UK
  3. 3Byers Eye Institute, Stanford University School of Medicine, Stanford, California, USA
  4. 4Ophthalmology, University of Virginia Health System, Charlottesville, Virginia, USA
  5. 5Kresge Eye Institute, Wayne State University School of Medicine, Detroit, Michigan, USA
  6. 6DoctorGoggle, Grand Cayman, Cayman Islands
  7. 7Legacy Devers Eye Institute at Legacy Good Samaritan Medical Center, Portland, Oregon, USA
  8. 8King's College London, London, UK
  1. Correspondence to Dr Edward Korot, Beaumont Eye Institute, Beaumont Health System, Royal Oak, Michigan, USA; ekorot{at}

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Summary box

What are the new findings?

  • We highlight the need for purpose-built, head-mounted displays tailored to the specific demands of ophthalmic surgery.

  • Our paper demonstrates the potential of using head-mounted displays in ophthalmic microsurgery.

How might it impact on healthcare in the future?

  • The widespread implementation of these devices will help alleviate musculoskeletal pain commonly seen in ophthalmologists who operate with traditional microscope.

  • Digital visualisation devices will allow for enhanced visualisation of the surgical field and digital overlays not attainable with a traditional analogue visualisation.

  • We demonstrate a collaborative iterative device design approach between clinician users and device manufacturers.


Surgical microscopes are necessary for the performance of a majority of ophthalmic procedures. However, the use of traditional microscopes often leads to abnormal body positioning, musculoskeletal pain and fatigue.1–3 Furthermore, pain in the back, neck and shoulder is disproportionately reported among ophthalmologists as compared with other medical specialties. The resulting injuries sustained may lead to chronic pain and decreased career longevity.4 Increasing use of three-dimensional (3D) heads-up display (HUD) monitors has been reported for both anterior and posterior segment ophthalmic procedures, as well as in several other surgical fields.1 HUD implies visualisation through a digital 3D display panel rather than through a traditional operating microscope.5 These systems, such as the TrueVision 3D Visualization System (TrueVision Systems, Santa Barbara, California, USA), use glasses with polarised lenses in conjunction with a consumer 3D monitor to create a stereoscopic image.6 7 Several use cases for image overlays have been explored, including image-guided toric intraocular lens positioning during cataract surgery and intraoperative ocular coherence tomography for posterior segment procedures.1 8 HUDs have been reported to decrease surgeon fatigue by allowing ergonomic body posturing without compromising visualisation, technical difficulty, surgical time, visual outcomes or complication rates.5 9 10 Additional benefits of moving from optical to digital visualisation include high-resolution digital zoom, increased depth of focus and decreased retinal light exposure through postprocessing techniques allowing lower illumination levels. This decreases the risk of phototoxicity as compared with traditional microscopes.11–14 Furthermore, these systems also allow for enhanced teaching potential over traditional microscopes, with trainees having visual fidelity mirroring that of the primary surgeon, and have demonstrated relatively short learning curves.15–17 As is the case with any novel technology, HUDs are not without their drawbacks. Headaches, nausea, visual disturbances and discomfort caused by body positioning necessary to view the display around the microscope have all been reported; blockage of anaesthesiologist’s patient access due to the large monitor size and necessary location has been a frequent complaint.18

A proposed alternative digital surgical visualisation method to HUD is head-mounted displays (HMDs). HMDs consist of a device with a microdisplay panel for each eye mounted on the surgeon’s head. HMDs are then connected to a 3D camera attached to the microscope optics. This allows displaying separate channels of low latency stereoscopic video to each eye. Due to their digital format, HMDs provide similar benefits to traditional microscopes as HUDs, including improved ergonomics, decreased illumination levels and ability to apply digital overlays during surgery. However, the unique advantages of HMDs include decreasing reliance on bulky monitors, no loss in image quality or stereopsis with changes in surgeon positioning, and less strain from decreased body position maintenance requirements as compared with HUDs.5 19 Various HMDs have been evaluated for surgical use, including the HMS-3000MT (Sony, Tokyo, Japan) and Clarity platforms (Beyeonics, Haifa, Israel).1 7 The former has been reported to incorporate improved ergonomics, high image quality, depth perception and spatial orientation with a 45o diagonal field of view and short learning curve.6 7 The Clarity platform pilot study found it to have comparable image quality with that of standard operating microscopes with improved magnification, half the required light, comfortable posture, intuitive head motions and no reported fatigue. However, the Clarity platform is not wireless, which may compromise ergonomics and functionality.20

In our prior pilot study, we introduced the use of the Glyph (Avegant, San Mateo, California), a head-mounted virtual retinal projection display, to perform vitreoretinal surgery.18 Initial data suggested that it provided exceptional visualisation, with high depth of field and lower required illumination levels as compared with the TrueVision HUD. In that trial, most surgeons reported a high level of confidence in safely performing procedures with the device. Additionally, with its small physical size, the device provided a novel simultaneous intraocular and extraocular visualisation, potentially useful for scleral depression during vitreoretinal procedures. Unfortunately, several shortcomings were reported, including lack of ear and nose comfort, poor contrast and an inability to simultaneously wear spectacles with the device. Vitreoretinal surgery entails unique requirements for digital visualisation systems, including high contrast ratios for manipulation of transparent tissues such as the vitreous and retina. Additionally, the relatively long operating times necessitate exceedingly high comfort while wearing a head-mounted device. To address the limitations of other HMDs within these constraints, we developed an HMD purpose-built for ophthalmic microsurgery. Our primary goals included improved visualisation, comfort for extended wear times, ability to wear spectacle correction and enabling surgeons to look around the HMD for simultaneous intraocular and extraocular surgical field visualisation. We report objective microsurgical skill performance of a modified Avegant Glyph device (HMD2), as well as performance using our purpose-built micro-organic light-emitting diode (OLED) HMD prototype, termed HMD3.

Methods and prototype design

HMD2 microsurgical performance study

The Glyph device from our prior feasibility study was augmented by opening the ear cuffs to allow for better ventilation and hearing in the operating room. With this prototype, termed HMD2, and a standard operating microscope, we compared time spent on a microsurgical corneal suturing task. We timed eight ophthalmologists, six of whom were surgeons in training. The task consisted of performing four corneal sutures with 10–0 nylon tied with ‘3-1-1’ interrupted square knots on iatrogenically induced lacerations of porcine corneas. This task was performed up to two times by each surgeon via standard ophthalmic operating microscope, on which all participants had prior operating experience, and two times with the HMD2 connected to a TrueVision 3D camera attached to the microscope. The participants were timed from first needle pierce to completion of the fourth suture knot.

HMD3 prototype design and development

An HMD3 custom prototype was developed to address several of the shortcomings identified in previous tests (figure 1). The HMD3 was designed using two Sony OLED panels with a 0.5″ diagonal with a resolution of 1280 by 960 pixels each. OLED displays provide the highest contrast ratio and were selected for this reason. Display elements were assembled by individually mounting the OLED panels into optical housings with four element optical lenses. The display elements have a field of view of 38°, an effective focal distance of approximately 2 m and eye relief of 45 mm allowing the use of corrective lenses. The two display modules are connected together in an optical assembly with a vergence of approximately 2 m and independent interpupillary distance adjustment. The entire optical module mounts to a standard binocular indirect ophthalmoscopy mount and displays 3D content from a high-definition multimedia interface (HDMI) input source such as a 3D camera connected to an operating microscope. To reduce wiring complexity in the operating room, power is drawn directly from the HDMI cable for a one-cable design.

Figure 1

Prototype head-mounted display. (A) Display unit containing micro-organic light-emitting diode display, lenses and electronics. (B) Combined power and high-definition multimedia interface display cable. (C) Head mount based on binocular indirect ophthalmoscope.

HMD3 microsurgical performance study

We compared time spent on a microsurgical suturing task completed while wearing the HMD3 and while using a standard operating microscope. The surgeons each performed four sutures with 9–0 nylon tied with ‘3-1-1’ interrupted square knots on an iatrogenically induced laceration of a latex suture practice board. We timed seven ophthalmologists, five of whom were surgeons in training. This task was performed up to two times by each surgeon via standard ophthalmic operating microscope, on which all participants had prior operating experience, and two times with the HMD3 connected to a TrueVision 3D camera attached to the microscope. The surgeons were timed from the first needle pierce to completion of the fourth suture knot. Surgeons then filled out a qualitative survey, which was scored on a 5-point scale (figure 2, table 1).

Figure 2

HMD3 surgeon survey results. Depth of field and confidence scored according to a 5-point scale on the x-axis. Communication scored on a scale between 0=‘severely hinders communication’ and 5=’does not hinder communication at all’. *Attending-level surgeon. 3D, three-dimensional; HMD3, head-mounted display 3; OR, operating room.

Table 1

Modified Likert survey scale for participants of the HMD3 trial

Statistical analysis

GraphPad Prism V.6 (San Diego, California) was used for statistical analysis. Two-tailed Wilcoxon matched-pairs signed-rank test was performed to compare surgeon performance between HMD and microscope for each HMD trial, and 95% CI was used where calculated. When tasks were attempted twice, the second attempt was used for paired comparison between visualisation methods in order to be more representative of participants’ performance after initial adaptation. Results are reported as mean (95% CI) number of minutes to complete the microsurgical task. Statistical level of significance was set at an alpha of 0.05.


The mean (95% CI) number of minutes to complete the microsurgical task was higher in surgeons using the HMD2 (6.20 (4.80 to 7.60)) as compared with the operating microscope (5.04 (3.96 to 6.13)) (figure 3). The difference (95% CI) between the means was 1.16 (0.62 to 1.70) min (p=0.008). The mean number of minutes to complete the microsurgical task was higher in surgeons using the HMD3 (6.24 (5.04 to 7.44)) as compared with the operating microscope (5.29 (4.22 to 6.36)) (figure 3, table 2). The difference between the means was 0.96 (−0.02 to 1.93) min (p=0.11).

Figure 3

Comparison of surgical task performance between head-mounted displays (HMDs) and surgical microscope. (A) Time in minutes to place four sutures in the latex suture board using HMD3. (B) Time in minutes to place four corneal sutures in the porcine eyes with HMD2. The lines represent mean with error bars at 95% CI. The mean (95% CI) number of minutes to complete the microsurgical task was higher in surgeons using the HMD2 (6.20 (4.80 to 7.60)) as compared with the operating microscope (5.04 (3.96 to 6.13)). The difference (95% CI) between means was 1.16 (0.62 to 1.70) min (p=0.008). The mean (95% CI) number of minutes to complete the microsurgical task was higher in surgeons using the HMD3 (6.24 (5.04 to 7.44)) as compared with the operating microscope (5.29 (4.22 to 6.36)). The difference (95% CI) between means was 0.96 (−0.02 to 1.93) min, although not statistically significant (p=0.11).

Table 2

Prototype HMD3 versus conventional microscope results

Of the eight surgeons who performed the HMD2 task twice, the average time to complete the task improved from the first to the second attempt by 16.9% (1.28 min, p=0.015) (figure 4). Of the two surgeons who performed the HMD3 task twice, the average total time improvement between the first and the second attempt of HMD3 was 18.9% (1.32 min). Conversely, of the two surgeons who completed the microscope task twice, the second microscope trial was on average 4.6% slower (0.55 min) (table 2). This aligns with expectations of improvement on the novel HMD2 and HMD3 devices with successive trials due to increasing familiarity, and suggests potential for continued improvement with further practice.

Figure 4

Difference in minutes between trials 1 and 2 of HMD2 as compared with best microscope time. Each point is a participant. All eight participants completed two trials of HMD2. Three participants completed two trials of the microscope (due to time constraints). The best microscope time is used for both comparisons. Of the eight surgeons who performed the HMD2 task twice, the average time to complete the task improved from the first to the second attempt by 16.9% (1.28 min, p=0.015). HMD2, head-mounted display 2.

Surgeon survey free-text responses commented on qualitative improvement of HMD3 as compared with HMD2 in ergonomics, field of view and vertex distance, which enabled those with refractive error to wear their spectacles simultaneously with the device. Adequate to high confidence was reported in performing surgical procedures with HMD3 (mean (SD) confidence: 3.63 (0.744), on a 5-point modified Likert scale, with 1=very low, 2=low, 3=adequate, 4=high and 5=very high) (figure 2). One surgeon reported pixilation artefacts on the needle, and one reported that the unit was too heavy. Surgeons further reported a lack of ‘screen door’ pixilation effect, which represents high pixel fill, on HMD3, and the majority reported high depth of field with no perceptible latency.


We present a study using objective measures to compare microsurgical task performance with a standard microscope and two different types of 3D HMDs, one of which we purpose-built for ophthalmic microsurgery. HUDs have been increasingly used for ophthalmic surgery to improve ergonomics and visualisation, with good surgical outcomes reported. However, several drawbacks for these systems persist, including bulky equipment size and residual issues with surgeon body positioning.1 15–18 We believe these may be addressed with HMDs.

Commercial HMDs have been studied for use in ophthalmic microsurgery in order to overcome the drawbacks of HUDs. However, adoption of these devices has been hindered by difficulty of use with spectacles, ergonomics and obstruction of the surgeon’s peripheral visual fields.1 4 5 7 18 19 The unique constraints of vitreoretinal and ophthalmic surgery, including high contrast ratios for manipulating transparent membranes, led us to develop an OLED-based HMD3 designed specifically for ophthalmic microsurgery.

Surgeons indicated qualitative improvement with HMD3 over an adapted consumer device (HMD2). The participants noted gains with regard to field of view, frame rate, contrast and comfort (figure 5). An improvement in subjective image latency was noted by participants in the study from HMD2 to HMD3, with lag being described as imperceptible in HMD3. Several limitations with the device still exist; a minority of surgeons (2 out of 8) noted issues with stereopsis, depth of field, resolution and weight of the device, all of which we aim to address with successive prototype iterations. Nevertheless, adequate to high confidence was reported in performing surgical procedures with the HMD3 after relatively little time with the device.

Figure 5

Photograph of a researcher wearing the HMD3 in an experimental set-up. The device is connected to the operating microscope with a three-dimensional camera. HMD3, head-mounted display 3.

Limitations of the study include differing tasks for evaluating HMD2 and HMD3 (porcine corneal vs latex laceration repair) and differing participant pools due to trainee graduation between trials, with five of our eight surgeons completing both device trials. Surveys were added as a component of the HMD3 trial, and not administered as part of the earlier HMD2 trial. Similarly, quantitative performance on a per-participant level could not be directly compared for the two HMDs. HMD2 and HMD3 performance was therefore evaluated independently in relation to baseline operating microscope. The trials were performed by majority trainee-level ophthalmologists; preferably an evaluation would have used a larger proportion of attending-level anterior and posterior segment ophthalmologists, as they are the device’s target users. These limitations were encountered due to equipment and surgeon availability constraints, respectively. Time constraints also prevented a repeat attempt of the HMD3 versus microscope task in all but two surgeons for each visualisation method. Although we performed a survey for HMD3 evaluation, it was not anonymous, potentially biasing participants. The latex platform has the advantage of having the ability to perform evaluation in a live operating room, where porcine specimens are prohibited. Although surgeons performing the HMD3 task spent close to 1 min longer as compared with the microscope, given they are more experienced in the use of the latter, HMD3 efficiency is expected to improve with practice. We observed a trend towards decreased time on the second task attempt during our HMD2 trial (figure 4) and expect a similar trend with HMD3 in trials and real-world use. In addition, participants with greater surgical experience tended to require a shorter amount of time to complete the surgical task while using HMD3. This indicates the translatability of operating microscope skills to the HMD3 device and the potential for a smooth transition to using the new device for surgeons experienced with a traditional operating microscope.

Although the Food and Drug Administration (FDA) and CONSORT (Consolidated Standards of Reporting Trials) groups provide guidance for non-inferiority trials for the pharmaceutical industry, limited guidance exists for objective medical device metrics.21–23 FDA drug trial guidance suggests non-inferiority be defined as the state where the difference between mean effect sizes is less than or equal to 50% of the upper 95% CI for the control intervention.21 22 Such trials are heavily influenced by the non-inferiority margin, which agencies recommend is set from ‘a combination of statistical reasoning and clinical judgement’, but analyses show the reasoning for setting such margins is absent in up to 50% of trial reporting.24 Our study represents the need for objective performance-based guidance from regulatory agencies for novel device performance.

In construction of the HMD3 device, we accomplished the optically challenging goal of simultaneously increasing vertex distance to allow wearing of spectacles, while maintaining image field of view and improving comfort. As a compelling cumulative metric of surgeons’ impression with the HMD3, all participants reported at least adequate confidence in performing the procedures, despite their limited experience with the device. Further feasibility testing with more complicated extraocular and intraocular manoeuvres is warranted beyond our representative suturing task. In particular, we feel that our device has potential for application in posterior segment surgery and plan to incorporate full simulations of these procedures in future studies. Furthermore, successive prototypes will be necessary for both regulatory approvals and adoption to occur.25 Addressing these challenges iteratively with incorporation of direct feedback from ophthalmologists demonstrates the benefit of an HMD which is purpose-built for ophthalmic surgery.


Maintaining high visualisation fidelity while enabling adequate ergonomics to avoid injury remains a challenge in ophthalmic surgery. Purpose-built HMDs show potential to address these issues, while providing novel advantages inherent to a digital visualisation system. In construction of the HMD3 device, we objectively demonstrate comparable ophthalmic skill performance. Further feasibility testing and iterative prototypes are warranted.

Ethics statements

Ethics approval

The device trials were exempt from review from institutional review board.


George Williams provided guidance and discussion and Natasha Conner provided provisioning of laboratory equipment for the study.



  • EK and MMR contributed equally.

  • Contributors EK contributed significantly to research design, data acquisition, research execution, data analysis and interpretation, and manuscript preparation. MMR contributed significantly to data acquisition, research execution, data analysis and interpretation, and manuscript preparation. AE contributed significantly to manuscript preparation. AT contributed significantly to research design and manuscript preparation. CB contributed significantly to manuscript preparation. EHW contributed significantly to data analysis and interpretation and manuscript preparation. MAR contributed significantly to research design, data acquisition and research execution.

  • Funding The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors.

  • Competing interests EK reports non-financial support from DoctorGoggle, during the conduct of the study. The evaluated device was provided by DoctorGoggle.

  • Provenance and peer review Not commissioned; externally peer reviewed.