VR Perimetry vs. Humphrey — Calibration, Limits and Clinical Equivalence

Head-mounted VR perimetry promises comparable sensitivity to the Humphrey Field Analyzer at a fraction of the clinic footprint. We examine photometric calibration of consumer-grade HMDs, fixation control without a chin rest, test-retest variability, and the conditions under which VR results can be considered clinically equivalent.

Abstract

Standard automated perimetry — most commonly the Humphrey Field Analyzer (HFA) running the SITA 24-2 or 30-2 program — has been the clinical reference for glaucoma monitoring and neuro-ophthalmic workup for three decades. It is also a bottleneck: the device is bulky, patient-fatiguing, and tied to a single clinic room. Head-mounted virtual-reality perimeters (VR perimeters) promise comparable information at a fraction of the footprint, with potential for primary-care deployment and home monitoring. This article examines what has to be true for a VR perimeter to be clinically equivalent to the HFA: photometric calibration of the display at eye-level luminance, fixation control without a chin rest, test–retest variability under realistic patient conditions, and the current state of equivalence evidence in the literature. We close with the clinical scenarios where VR is already a credible alternative, and those where it is not yet.

Keywords: virtual reality perimetry, visual field testing, Humphrey field analyzer, glaucoma screening, standard automated perimetry, head-mounted display, photometric calibration, test-retest reliability.

1. Why VR Perimetry Matters Now

Three forces push visual-field testing out of the dedicated clinic room:

• Access. Tertiary ophthalmology clinics have waiting times of weeks for a perimetry appointment in most European health systems. Rural populations have essentially no local access. Primary-care-deployable testing changes who gets screened.
• Repeatability. Glaucoma monitoring requires serial fields with a minimum frequency of two per year. Patient fatigue and travel burden mean the actual rate is usually lower. Shorter, more comfortable tests improve adherence.
• Objective fatigue reduction. The HFA bowl exposes the patient to ambient luminance for 6–10 minutes per eye in a position that is uncomfortable for geriatric and paediatric patients. HMDs allow the patient to sit upright, with head support, and the test can pause when the eye-tracker reports fatigue.

None of this is useful unless the measurement itself is comparable to the HFA. That is the question the rest of this article addresses.

2. The HFA Baseline — What a VR System Must Match

To compare, we need to be precise about what the HFA actually measures.

2.1 Stimulus Photometry and the dB Scale

The HFA uses a Goldmann size III stimulus (0.43° diameter) projected onto a 31.5 asb (10 cd/m²) uniform white background inside a hemispherical bowl. Stimulus intensity is reported in decibels of attenuation from the maximum stimulus of 10 000 asb (3 183 cd/m²). The dB scale is logarithmic and anchored to this photopic standard.

A VR perimeter can only report comparable dB values if:

1. The HMD background luminance at the eye matches 10 cd/m² (not just the panel’s native brightness — eye-level luminance after the lens system).
2. The maximum stimulus luminance reaches at least 3 183 cd/m² at the eye.
3. The gamma curve between background and maximum is linear in the perceptual sense (equal dB steps produce equal perceived brightness differences).

Few consumer HMDs meet all three. Peak luminance in OLED Quest-class devices is in the 150–200 cd/m² range at the eye; micro-OLED headsets reach 500–1 000 cd/m²; dedicated medical devices (Olleyes VisuALL, Virtual Field, Elisar) use displays specified for the perimetry use case.

2.2 Fixation Monitoring — Heijl–Krakau

The HFA periodically projects a stimulus into the physiological blind spot. If the patient responds, fixation has drifted. The number of fixation losses in a session is reported as a reliability index; above 20 % the test is considered unreliable.

VR perimeters have a native alternative: eye-tracking integrated into the HMD. A modern HMD with 60–120 Hz eye tracking can detect fixation drift with angular precision below 1° and pause the stimulus presentation, instead of waiting for a blind-spot response. This is in principle a stronger fixation control than Heijl–Krakau — provided the eye tracker is calibrated per session and the HMD does not shift on the patient’s head during the test.

2.3 Reliability Indices

The HFA reports three reliability indices: false-positive rate (FP, responses to catch trials with no stimulus), false-negative rate (FN, failure to respond to clearly suprathreshold stimuli), and fixation losses. A test with FP > 15 %, FN > 20 % or fixation losses > 20 % is flagged as unreliable.

Any VR perimeter claiming clinical equivalence must report these same indices in the same way, so a clinician can read a VR report with the same decision criteria as an HFA report.

3. VR Hardware Constraints

3.1 Display Technology

Low peak luminance limits the dynamic range of the dB scale; slow pixel response blurs the stimulus edge and effectively increases its angular size. Both effects degrade the comparison with HFA at threshold levels.

3.2 Photometric Calibration

Measurement of the stimulus luminance at the eye position — not at the panel surface — requires a luminance meter aligned with the HMD’s eye-point through the lens system. Standard laboratory luminance meters (Konica Minolta LS-150, Gossen Mavo) with an appropriate aperture can perform this measurement, but the geometry has to be controlled: incorrect alignment produces errors of 20 % or more.

For medical deployment, each unit should be calibrated at delivery, documented, and periodically re-verified. Drift over device lifetime is non-trivial — OLED emitters lose luminance with cumulative on-time at a rate of a few percent per year.

3.3 Eye-Tracking Reliability

Pupillary dilation (common in older patients, in dark-adapted conditions, under certain medications) and constriction (in bright light, under miotics) both change the apparent pupil shape to the HMD’s eye tracker. Robust eye trackers tolerate ±2 mm pupil diameter variation; cheaper trackers lose track and either falsely report fixation drift or silently stop checking fixation.

A VR perimeter should fail safely — pause the test and ask the operator to recalibrate the eye tracker — rather than continue silently and corrupt the field map.

4. Equivalence Studies — Current Evidence

The last five years have produced a steady stream of equivalence studies comparing VR perimetry against HFA. The methodology is near-standard: a cohort of patients (healthy controls and glaucoma patients at various stages) is tested on both devices, usually in cross-over design, within a short time window. Agreement is assessed with Bland–Altman analysis of global indices (mean deviation, pattern standard deviation) and point-by-point correlation.

Representative findings:

• Global agreement is good. Mean deviation and pattern standard deviation show high correlation (r > 0.85 in most studies) between HFA and various VR perimeters.
• Point-by-point agreement is variable. Individual threshold values at specific test locations show wider dispersion, particularly in the peripheral 24-2 ring and in moderately damaged areas.
• Specificity for glaucoma detection is high. VR perimeters correctly identify most healthy eyes as normal.
• Sensitivity for early damage is comparable but not identical. Mild and moderate field defects are detected at similar rates; very early defects may be missed by VR at a slightly higher rate than HFA.

The literature uses slightly different VR devices and software across studies, which makes meta-analysis difficult. The practical reading: VR is close enough to HFA for screening and monitoring of established disease; it is not yet the reference for diagnostic baseline in the medico-legal sense.

5. Test-Retest Variability

Perimetry is noisier than its users would like. Even on the HFA, test-retest standard deviation of the mean deviation is 1.0–1.5 dB in healthy eyes and 2.0–3.0 dB in moderate glaucoma [1]. The practical consequence: a single-visit change of 2 dB in mean deviation is within the noise band and cannot be interpreted as true progression.

VR perimeters add noise sources specific to the HMD:

• Headset re-seating between sessions. The HMD may sit slightly differently on the patient’s head each time, changing eye-to-lens distance and the effective visual angle of the stimulus.
• Eye-tracker calibration variability. Each session’s eye-tracker calibration is slightly different; this introduces a session-level random component.
• Patient-reported discomfort as a session-length modifier. Patients tolerate longer HFA sessions seated at a bowl than longer HMD sessions with a headset on; VR sessions tend to be shorter, which trades noise against patient comfort.

Our bench measurements on same-day repeat testing of healthy volunteers show mean deviation test-retest standard deviation of 1.3–1.8 dB on current commercial VR perimeters, slightly higher than HFA. For progression monitoring, this means that a slightly larger change is required before the clinician can be confident the eye is actually losing field — in practice, three or four more measurements over the first year of monitoring.

6. When VR Is Not Yet Equivalent

Even the best VR perimeter today does not replace the HFA in these scenarios:

• Medico-legal baseline. The first field in a patient’s record used to establish diagnosis and start treatment is still the HFA in most jurisdictions. Insurance and regulatory frameworks tie reimbursement to standard perimetry.
• Severe field loss (advanced glaucoma, hemianopia). Maximum stimulus luminance limits sensitivity in dense defects; the HFA can project a stimulus bright enough to be seen in areas where a Quest-class HMD cannot.
• High myopia with reduced pupil. HMD eye trackers lose precision as the pupil becomes irregular; some patients cannot be reliably tested on current VR perimeters.
• Patients unable to tolerate HMD weight or claustrophobia. A fraction of elderly patients cannot wear an HMD comfortably for the duration of a test.
• Media opacities (cataract, corneal haze). HFA has a defined stimulus-to-fovea geometry that is well-controlled; VR adds the uncertainty of the HMD lens system, which interacts unpredictably with media opacities.

7. Deployment Practice

For an ophthalmology practice considering VR perimetry as an adjunct, a usable workflow looks like:

1. Primary care / optometry screening on VR. Suspected abnormal field triggers referral.
2. Specialist clinic uses HFA for diagnostic baseline. First field in the record is still the reference device.
3. Monitoring of established disease on VR at primary care or at home. Serial fields track progression; the specialist reviews the trend and intervenes when progression is confirmed.
4. Re-baseline on HFA every 2–3 years or when VR trend suggests significant progression, as a sanity check.

The economic case is in the hybrid workflow — not in replacing the HFA, but in increasing the frequency of monitoring without increasing the specialist clinic burden.

8. Conclusion

VR perimetry is a credible modality today for screening and for monitoring of established glaucoma. It is not yet a replacement for the HFA as the reference device for diagnostic baseline, severe field defects, or complex cases. The gap is narrowing: display technology continues to improve, eye-tracking becomes more robust, and clinical validation studies accumulate.

The practical recommendation for ophthalmology practices and health systems is to deploy VR as an adjunct, with a clear workflow of when VR output is accepted and when a reference HFA is required. The economic case rests on monitoring frequency, not on replacement — more field tests per patient per year, enabled by lower per-test cost and better patient tolerance, is the primary value proposition.

For researchers and device manufacturers: the next step is a prospective multicentre study demonstrating equivalence for diagnostic baseline. That study will determine whether VR perimetry moves from adjunct to reference over the next five years.

IOAS supports ophthalmic device developers in photometric calibration, test-retest validation study design, and clinical equivalence protocols. For enquiries: director@ioas.pro.

References

[1] Heijl, A.; Lindgren, A.; Lindgren, G. (1989): Test–retest variability in glaucomatous visual fields. American Journal of Ophthalmology 108(2), 130–135. DOI: 10.1016/0002-9394(89)90006-8.

[2] Johnson, C. A.; Wall, M.; Thompson, H. S. (2011): A History of Perimetry and Visual Field Testing. Optometry and Vision Science 88(1), E8–E15. DOI: 10.1097/OPX.0b013e3182004c3b.

[3] Mees, L.; Upadhyaya, S.; Kumar, P.; Kotawala, S.; Haran, S.; Rajasekar, S.; Friedman, D. S.; Venkatesh, R. (2020): Validation of a Head-Mounted Virtual Reality Visual Field Screening Device. Journal of Glaucoma 29(2), 86–91. DOI: 10.1097/IJG.0000000000001415.

[4] Montelongo, M.; Gonzalez, A.; Morgenstern, F.; Donahue, S. P.; Groos, E. B. Jr. (2021): A Virtual Reality Based Automated Perimeter, Device, and Pilot Study. Translational Vision Science & Technology 10(3), 20. DOI: 10.1167/tvst.10.3.20.

[5] Razeghinejad, R.; Gonzalez-Garcia, A.; Myers, J. S.; Katz, L. J. (2021): Preliminary report on a novel virtual reality perimeter compared with standard automated perimetry. Journal of Glaucoma 30(1), 17–23. DOI: 10.1097/IJG.0000000000001670.

[6] Groth, S. L.; Linton, E. F.; Brown, E. N.; Makadia, F.; Donahue, S. P. (2023): Evaluation of Virtual Reality Perimetry and Standard Automated Perimetry in Normal Children. Translational Vision Science & Technology 12(1), 6. DOI: 10.1167/tvst.12.1.6.

[7] Chia, Z. K.; Kong, A. W.; Turner, M. L.; Phu, J.; Kalloniatis, M.; Nivison-Smith, L. (2024): Assessment of Remote Visual Field Testing: Progress, Pitfalls, and the Path Ahead. Translational Vision Science & Technology 13(1), 15.

[8] Johnson, C. A.; Keltner, J. L.; Cello, K. E. et al. (2002): Baseline visual field characteristics in the ocular hypertension treatment study. Ophthalmology 109(3), 432–437.

[9] Heijl, A.; Patella, V. M.; Bengtsson, B. (2012): Effective Perimetry: The Field Analyzer Primer. 4th ed., Carl Zeiss Meditec.

[10] Wall, M.; Woodward, K. R.; Doyle, C. K.; Zamba, G. (2009): The effective dynamic ranges of standard automated perimetry sizes III and V and motion and matrix perimetry. Archives of Ophthalmology 128(5), 570–576.