A new study published in Frontiers in Ophthalmology offers a fresh answer to that access problem. Researchers led by George Y. X. Kong of the University of Melbourne and the Centre for Eye Research Australia, together with colleagues at Marengo Asia Hospitals in Gurugram, India, and the Mayo Clinic in Jacksonville, Florida, compared a browser-based perimetry platform called Melbourne Rapid Fields-Web, or MRF-Web, against the Octopus 600, a widely used conventional clinical perimeter from Haag-Streit Diagnostics. The question was simple but consequential: can a test that runs on an ordinary laptop screen, in a web browser, produce results close enough to a dedicated clinical device to matter for patient care?

The study included 113 patients undergoing routine visual field assessment for glaucoma or suspected glaucoma at a single center in India. All participants had visual acuity of 6/12 (20/40) or better, and eyes with extreme refractive errors or recent intraocular surgery were excluded. The cohort was heavily weighted toward early disease: based on the Hodapp-Parrish-Anderson classification applied to the Octopus 600 Mean Deviation value, 103 of the 113 participants had mild visual field loss, while only five fell into the moderate category and five into the severe category. The average age was 56.5 years, with 63 men and 50 women taking part. Where both eyes were tested, a single study eye was selected for the main summaries to avoid statistical complications from correlated measurements within the same person.

The technical setup of the MRF-Web test reveals how much engineering goes into making a laptop screen behave like a clinical perimeter. Testing used a 24-2 stimulus grid, the same spatial layout familiar from standard automated perimetry, with sensitivity thresholds estimated through a rapid neighborhood-based Zippy Estimation by Sequential Testing (ZEST) approach using seven discrete threshold steps across a 0 to 30 dB range. Crucially, the size of the light stimuli scales with eccentricity, growing larger toward the periphery of the visual field. This design choice reflects the physiology of spatial summation in the retina and serves two purposes: it reduces measurement variability and it maintains a usable dynamic range in damaged fields, where small stimuli would otherwise frequently fall below the detection floor. Previous work by the same group has shown that scaling stimulus size with eccentricity returns approximately constant thresholds across the field and that larger stimuli reduce variability in eyes with glaucomatous damage without sacrificing detection capability.

Because a web browser can run on almost any computer, the researchers had to control for the one variable that consumer hardware introduces: the display itself. Before every test, a clinical assistant calibrated the 14-inch laptop using an in-software routine in which the participant adjusts an inverted U-shaped virtual ruler to match the width of a standard credit card, allowing the software to calculate the screen’s physical dimensions and resolution. Screen brightness was set to maximum at 350 candela per square meter, the viewing distance was fixed at 33 centimeters, and the fellow eye was occluded. Testing took place in a darkened room with the participant’s habitual near correction in place and natural pupils. Perhaps the most striking feature is the software’s use of the laptop’s webcam for facial tracking: the camera continuously monitors viewing distance by comparing the current size of the patient’s face to the size recorded at the start, triggering a moving animation at the fixation target if distance drifts by more than 5 percent and a voice prompt to readjust at 10 percent. Fixation was monitored with standard blind-spot techniques, and gaze stability was plotted on the final report. All voice guidance was delivered in Hindi, a deliberate choice to remove language barriers in the local population.

The headline result concerns agreement between the two devices on Mean Deviation, the global index that summarizes overall visual field loss. MRF-Web showed strong group-level concordance with the Octopus 600, with an intraclass correlation coefficient of 0.88 (95 percent confidence interval 0.81 to 0.92). Bland-Altman analysis, which examines how the differences between two measurement methods behave across the range of values, revealed a small systematic bias of -0.78 dB, meaning MRF-Web tended to return slightly more negative, that is, slightly worse, MD values than the Octopus 600. The 95 percent limits of agreement stretched from -4.72 to +3.17 dB, a range wide enough that the authors caution the two platforms should not be regarded as directly interchangeable for individual measurements. In practical terms, a patient’s MD value on one device cannot simply be compared against a value recorded on the other; serial monitoring should stay on a single platform.

For the pattern-based indices, which capture localized rather than generalized field loss, concordance was moderate rather than strong: an intraclass correlation coefficient of 0.72 (95 percent confidence interval 0.58 to 0.81) with a bias of -0.59 dB and limits of agreement from -4.41 to +3.24 dB. The authors are careful to explain why this comparison is inherently imperfect. MRF-Web uses its own neighborhood-based ZEST thresholding, eccentricity-scaled stimuli and a normative framework derived from 17,390 thresholds in normal and glaucomatous participants, whereas the Octopus 600 output was generated with the Dynamic thresholding strategy and converted to HFA-style indices using proprietary manufacturer software whose transformation formulas are not publicly available. The two pattern indices are therefore clinically analogous summary measures rather than mathematically identical quantities, and the study quantifies concordance between related device-specific measures rather than proving numerical equivalence.

Repeatability was assessed in a subset of 51 patients who underwent repeat MRF-Web testing within three months, 48 of them during the same session after roughly a ten-minute rest. Test-retest concordance was good, with an intraclass correlation coefficient of 0.86 (95 percent confidence interval 0.69 to 0.94) and limits of agreement from -3.57 to +5.20 dB. An intriguing pattern emerged: the difference between the two tests was significantly associated with the mean MD of the pair (Pearson r = -0.53, p < 0.001), with eyes showing greater field loss tending to score better on the second test. This may reflect a learning effect or severity-dependent variability, but with only seven participants in the repeatability subgroup having moderate or severe loss, the data cannot support reliable severity-stratified estimates. The authors are explicit that these findings support repeatability at a group level but do not yet establish suitability for longitudinal progression monitoring, the most demanding task in glaucoma care.

Where MRF-Web decisively outperformed its conventional rival was speed. The average per-eye test duration was 232.9 seconds, just under four minutes, compared with 342.4 seconds, nearly six minutes, on the Octopus 600, a saving of approximately 1.8 minutes per eye. In a busy clinic performing dozens of visual field tests a day, that difference compounds into meaningful gains in patient throughput and comfort. The trade-off appeared in reliability metrics: false-positive error rates were higher with MRF-Web at 4.9 percent versus 0.78 percent on the Octopus, a statistically significant difference, while false-negative rates were essentially equivalent at 4.5 and 4.1 percent. A comparison of fixation losses was not possible because the Octopus 600 does not record them. Mean MD values also differed slightly between devices, at -2.51 dB for MRF-Web versus -1.75 dB for the Octopus, consistent with the small systematic bias seen in the agreement analysis.

The study’s limitations are candidly acknowledged and they matter for how the results should be used. The single-center retrospective design, the overwhelming predominance of mild disease, the absence of repeat Octopus testing for direct comparison of device repeatability, and the use of a single calibrated laptop in one supervised environment all constrain generalizability. Performance across different screen types, languages, and unsupervised home settings was not assessed here, although prior studies of the MRF platform, including comparisons with the Humphrey Field Analyzer in multiple populations and a twelve-month home telemedicine study, suggest the approach travels well. Still, the core message is a significant one for global eye health: a browser-based perimeter running on a calibrated consumer laptop, guided by voice prompts in the patient’s own language, tracked a clinical-grade device closely at the group level while cutting test time by roughly a third. For the vast populations where dedicated perimeters are scarce, that combination of accessibility, speed and reasonable agreement could bring systematic visual field monitoring within reach, provided clinicians remember that numbers from different platforms are not interchangeable and that moderate and advanced glaucoma still await their own validation data.

Subject of Research: Comparison of a web-based perimetry platform with a conventional perimeter for glaucoma visual field testing

Article Title: Comparison of the Melbourne rapid fields-web device-independent online perimeter and the octopus 600 perimeter in glaucoma patients

Article References: Kong, G. Y. X., Vingrys, A., Ambawat, M., Mahboob, M., Bano, S., Dorairaj, S., & Bhartiya, S. (2026). Comparison of the Melbourne rapid fields-web device-independent online perimeter and the octopus 600 perimeter in glaucoma patients. Frontiers in Ophthalmology, 6, Article 1961375. https://doi.org/10.3389/fopht.2026.1961375

Image Credits: AI Generated

DOI: 10.3389/fopht.2026.1961375

Keywords: glaucoma, perimetry, visual field testing, Melbourne Rapid Fields, Octopus 600, telemedicine, web-based diagnostics, Mean Deviation, test-retest repeatability, Bland-Altman analysis, ophthalmology, digital health