Click to enlarge

Problems with the coordination of our eye movements can reveal important information about how well different parts of the brain are working. One such problem, called internuclear ophthalmoplegia (INO), is common in people with multiple sclerosis (MS). INO can cause double vision, especially when looking to the side, and may also lead to dizziness or noticeable compensatory head movements as patients try to keep vision stable. INO is caused by a disruption of brainstem pathways that coordinate both eyes. The responsible lesion can be very small and may be difficult to recognize on MRI. Even with careful clinical assessment and conventional testing, mild INO can be missed, particularly when findings are subtle.

Our work shows that a certified medical device combining virtual reality with eye tracking² can quantify tiny left–right differences during quick eye movements. Using these measurements, our approach has the potential to enable automated detection of patterns consistent with INO. This could support earlier recognition of subtle brainstem involvement and provide an objective measure to complement bedside exams and MRI in MS.

Click to enlarge

Objective:
The development of artificial intelligence (AI) models for diagnosing Parkinson’s disease (PD) using ocular motor parameters obtained from a virtual reality (VR)-based medical device.

Background:
PD diagnosis currently relies largely on clinical examination, underscoring the need for objective biomarkers to enhance diagnostic accuracy, early detection and disease monitoring. Eye movement and pupil function abnormalities have emerged as promising candidates, reflecting PD-related dysfunction in ocular motor control pathways.

Design/Methods:
Eighty-one patients with idiopathic PD were enrolled in Zurich (Switzerland) and Exeter (UK). Each participant underwent two visits, including standardized ocular motor testing with the VR-based medical eye tracker and corresponding manual examination. Data with poor tracking quality or excessive signal loss were excluded. The present analysis included 132 PD examinations and 148 healthy control examinations from an independent dataset. Each PD patient was matched to one control of similar age and sex using a globally optimal matching algorithm. For each parameter, standardized mean differences (SMDs) and Welch’s tests were computed. A supervised machine learning model was trained to classify PD versus controls based on the most discriminative parameters. Site-specific analyses were conducted for Zurich and Exeter.

Results:
Parameters showing the largest PD–control differences included saccades (accuracy, velocity, main sequence), vergence, ocular alignment, and pupil dynamics (constriction/dilation velocity and latency). The machine learning model achieved strong discrimination: Zurich—AUC 0.96 (F1: 0.92, sensitivity: 0.96, specificity: 0.89); Exeter—AUC 0.91 (F1: 0.84, sensitivity: 0.89, specificity: 0.79).

Conclusions:
Preliminary findings demonstrate that quantitative ocular motor and pupillary metrics can robustly distinguish PD from healthy individuals. Next steps include expanding the patient and control cohorts, merging datasets across sites, and correlating ocular motor parameters with clinical measures to refine diagnostic performance.

Click to enlarge

Introduction

• Diabetic Retinopathy (DR) causes retinal ischemia, leading to impaired dark adaptation due to high oxygen demand.

• Hypothesis: Recovery of the pupillary light reflex (PLR) is delayed in DR patients compared with healthy controls.

Methods

• Design: Multi-center, non-interventional pilot study comparing PLR recovery in DR patients with healthy controls (HC) and retinitis pigmentosa (RP) patients (positive controls).

   

• Procedure: Pupillometry examination using the PRET™ VR headset (machineMD AG, Bern, Switzerland)⁵.

   

• Stimulus protocol: Alternating bright and dark stimuli across 50 cycles (dark phase increased by 100ms per cycle) while recording pupil size.

   

• Analysis: Linear regression and a linear mixed-effects model were used to test whether pupil responses at different time points and changes in PLR amplitude over time differ between groups.

   

Results

• Analysis included 10 patients with DR, 10 patients with RP, and 11 healthy controls.

• In all groups, PLR amplitude increased with longer dark adaptation across the 50 stimulus cycles (Fig. 2).

• Adaptation dynamics (Fig. 2) and maximal PLR amplitudes (Fig. 3) were significantly lower in DR and RP patients.

Conclusion

• PLR amplitudes may enable detection of DR using a simple pupillary test with a commercial VR headset, likely reflecting oxygen deprivation in DR.

• PLR may serve as a simple diagnostic test for retinal dystrophies, similar to ERG, but easier to perform.

Purpose: Eye and pupil movements are critical biomarkers for assessing neurological conditions. Most functional eye exams are manually performed making them subjective and error-prone; contributing to an estimated 50% misdiagnosis referral rates for neuro-ophthalmic conditions. neos™ (machineMD, Switzerland) is a novel virtual reality (VR) based measurement device, utilizing a head-mounted VR display and stimuli with infra-red eye tracking and proprietary algorithms to perform neuro-ophthalmological based tests in a non-invasive, automated, and quantitative way. However, validation of eye-tracking-based tools like neos™ poses inherent challenges, particularly in separating the device’s performance characteristics from those of the examinee. This study outlines a methodology to isolate and evaluate neos’s performance independent of external influences to ensure measurement reliability and robustness.

Methods: We evaluated neos’s gaze precision and pupil size measurement outputs independently using a customized eye-robot that simulates human eye movements and in human participants. We used test-retest reliability as a robust, calibration-independent surrogate metric of gaze precision of robot data that was further analyzed using intraclass correlation coefficients (ICCs), standard error of measurements (SEM), and Bland-Altman analysis. For human participants, we assessed pupil size by comparing neos’s eye tracker measurements and infrared images to visual light images calibrated with a physical scale, using the iris diameter as a reference.

Results:  In robotic trials, neos™ achieved excellent test-retest reliability for gaze precision (ICC > 0.99, SEM = 0.04), with high agreement between repeated measurements. Bland-Altman plots confirmed minimal bias and narrow variability across measurements, demonstrating robustness even in noisy conditions. Human trials revealed a mean pupil size difference of 0.38 ± 0.20 mm and iris size difference of 1.66 ± 0.38 mm between eye tracker and reference measurements.

Conclusions: The neos™ device has the potential to automate functional eye movement and pupil response measurements with high precision and minimal operator dependency, which can enhance neuro-ophthalmic diagnostic accuracy in clinical settings. Future studies will investigate its performance in detecting a range of specific biomarkers.

Background: Persistent Postural-Perceptual Dizziness (PPPD) is a common chronic vertigo disorder triggered by postural changes, motion, and complex visual environments. The diagnostic criteria defined by the Bárány Society in 2017 are entirely subjective, and no objective biomarkers are currently available. This makes accurate diagnosis challenging, despite the condition’s substantial impact on quality of life. Optokinetic after-nystagmus (OKAN) has recently been proposed as a potential objective marker for visual dependence and functional vestibular disorders, including PPPD. However, traditional OKAN measurement techniques require specialized full-field optokinetic drums and are not feasible for routine clinical practice. The aim of this study is to determine whether Virtual Reality (VR) can reliably induced and measure the OKAN, and whether VR-derived OKAN parameters can distinguish PPPD patients from healthy controls. 

Methods: In this prospective cross-sectional pilot study, twelve healthy participants and thirteen PPPD patients underwent VR-based optokinetic stimulation with integrated eye-tracking⁵. The optokinetic stimulus consisted of a 40°/s full-field visual motion pattern presented for 60 seconds. All patients received standard clinical audiovestibular testing (vHIT, VNG, audiometry) and MRI. Outcome measures included OKAN presence, OKAN time constant (TC, s), OKN nystagmus score (%), test acceptability questionnaires (Likert /5) and test wellbeing questionnaires (Likert /5). 

Results: PPPD patients demonstrated significantly lower OKAN TC (3.05s vs. 19.72s in controls, p = 0.003) and lower OKN nystagmus scores (70.21% vs. 77.03% in controls, p = 0.032). The OKAN TC showed a sensitivity of 79% and a specificity of 83%. Test acceptability and wellbeing scores did not differ significantly between groups and consistently met predefined minimal viable product (MVP) thresholds (wellbeing: 3.0 vs. 3.3, MVP>3; acceptability: 4.3 vs. 4.7, MVP >4; both p > 0.05). 

Conclusion: The pilot study demonstrated that VR-based OKAN measurement is technically feasible, well tolerated, and enables OKAN assessment without specialized rotation chairs or optokinetic drums. The significant differences in optokinetic parameters between PPPD patients and healthy controls suggest that VR-OKAN could serve as a promising objective diagnostic tool. Further validation in larger and longitudinal cohorts is warranted to determine its clinical potential.

Objective
To evaluate the diagnostic performance of the neos® device as an automated tool for sensorimotor assessment.

Background
Sensorimotor examination is a cornerstone of diplopia evaluation but requires specialized training. Many neurologists encounter patients with diplopia or ocular misalignment but lack the tools or expertise for sensorimotor assessment.
The neos® device (machineMD, Bern, Switzerland) automates components of the sensorimotor exam, offering the potential to extend diagnostic capabilities to non-specialists and enable remote evaluation.

Design/Methods
Patients presenting to the neuro-ophthalmology clinic with diplopia or related complaints between September and December 2024, along with consenting controls, underwent testing with neos®. Results from the neos® sensorimotor exam were compared with conventional alternate cover testing in primary gaze. Diagnostic accuracy was assessed by comparing the full clinical diagnosis from in-person evaluation with the diagnosis of a blinded neuro-ophthalmologist who reviewed only neos® data and a one-sentence vignette, simulating remote consultation.

Results
Thirty-five participants were enrolled (mean age 55.0 = 18.5 years; 40% male). Abnormal sensorimotor findings explaining diplopia were identified in 18 (51.4%) by standard exam and in 23 (65.7%) by neos®. The sensitivity and specificity of neos® for detecting sensorimotor abnormality were 88.9% and 58.8%, respectively. neos® measurements of deviation were within 5 prism diopters or 25% (whichever was larger) in 60.6% of horizontal and 75.8% of vertical cases.
In a simulated telemedicine context, neos® data enabled accurate diagnosis in 21 (60.0%) cases. It measured a vertical deviation greater than 1 prism diopter in 15 (65%) of cases with no vertical deviation on clinical exam.

Conclusions
The neos® device provides a promising means to obtain sensorimotor exam information, but may provide artifactual data or false vertical deviations. With refinement in measurement precision and clinician experience in data interpretation, neos® could facilitate broader access to accurate strabismus assessment.

Objective:
Describe the utility of virtual perimetry in workup  of patients with functional visual loss (FVL) .

Background:
Functional vision loss can be difficult to prove especially when deficits are seen on automated perimetry. Virtual reality (VR) perimetry has emerged as a promising tool for assessing visual function in individuals with FVL.

Design/Methods:
Retrospective review of 5 patients ages 19-60 years presenting with visual defects on automated perimetry due to functional visual loss.  The patients presented with varied visual symptoms including abnormal automated perimetry in contrast to intact confrontational perimetry and normal structural exam. Virtual [gaze-dependent] perimetry (neos, machineMD) was performed and compared to automated perimetry.

Results:
While standard automated perimetry revealed  varied defects,  VR perimetry documented largely normal visual fields. 2 patients had monocular deficits on perimetry and normal virtual perimetry; 2 patients with monocular deficit on automated perimetry showed a homonymous pattern on virtual;1 patients with right homonymous defects on automated perimetry had normal virtual perimetry.

Conclusions:
VR-based perimetry provides a novel tool for assessing visual function  and may be a valuable adjunct in evaluating patients with suspected functional visual loss. The usefulness of virtual perimetry in functional patients may be related to a binocular paradigm with monocular data acquisition.

Click to enlarge

Purpose
Concussion frequently disrupts visual pathways, making neuro-ophthalmic assessment critical for both diagnosis and monitoring recovery. The purpose of this case study is to describe what appears to be, after literature searches on PubMed and Google Scholar, one of the first reported cases of post-concussive Horner’s syndrome, as well as oculomotor dysfunction, and to illustrate its resolution using a virtual-reality (VR) neuro-ophthalmic diagnostic platform. 

Case Report
A 44-year-old man developed frontal-temporal headaches provoked by near-vision tasks—scrolling on phone, reading, and gaming—after a ball struck his face several months earlier. While the patient demonstrated normal best-corrected visual acuity, his initial examination revealed left-eye Horner’s syndrome (upper-lid ptosis, lower-lid “reverse” ptosis, anisocoria with dilation lag) and saccadic intrusions on near-point-of-convergence testing. Ocular health was otherwise unremarkable. 

VR-based neuro-ophthalmic testing showed a comitant eso deviation, left-eye dilation lag without relative afferent pupillary defect, saccadic inaccuracies and corrections, reduced saccadic velocity in down going saccades > up going saccades, saccadic intrusions and endgaze nystagmus during smooth pursuit, mildly inaccurate convergence beyond 5°, and limited divergence (≈1.5°). 

Six months later the clinical signs of Horner’s syndrome had completely resolved, along with some signs of oculomotor dysfunction. Repeat VR testing demonstrated normal pupillary function, more accurate saccades, fewer saccadic intrusions and less endgaze nystagmus on pursuits, improved convergence, and absent divergence fusion; vertical-saccade velocities were unchanged. 

Conclusion
This case demonstrates that Horner’s syndrome can occur as a transient sequela of concussion and may resolve within months. Serial assessment with a VR neuro-ophthalmic platform objectively captured both the presence and the resolution of sympathetic dysfunction as well as broader oculomotor recovery. This helps underscore the technology’s value in the diagnosis and monitoring of concussion and, by extension, more severe traumatic brain injury. 

Click to enlarge

Introduction:  Virtual reality (VR) eye-tracking headsets are increasingly used in neuro-ophthalmology, allowing precise measurements in a controlled environment. This study evaluates their performance in measuring gaze direction and pupil size in children.

Methods:  For this study (swissethics-BASEC project-ID: 2023-D0055) 22 healthy children (12 female, 10 male; 3-12 years) and 145 adults and adolescents (76 female, 69 male, 13-92 years) were assessed with the VR- based medical eye-tracking device (neos™). The device examines oculomotor parameters by measuring pupil size and gaze directions. Descriptive statistics were conducted comparing the measurement precision in children with adolescents and adults.

Results:  The precision of gaze direction was lower in children (median 0.97°, 95th percentile 3.81°) compared with adolescents and adults (median 0.35°, 95th percentile 0.84°). Pupil size measurements were similarly precise in children (median 0.09 mm, 95th percentile 0.18 mm) compared with adolescents and adults (median 0.05 mm, 95th percentile 0.16 mm). Unstable eye-tracking and low task-compliance were more frequently observed in children than in adults and adolescents.

Conclusion/Relevance:  The VR-based neos™ eye-tracking system can be utilized for precise assessments of pupillary function in children, whereas the assessments relying on eye-gaze should be interpreted with caution. The measurement precision in children could further be improved by adapting the hardware to fit smaller heads better and by optimising the stimuli for children. Preliminary data in 20 children (with Galactosemia and healthy controls; swissethics-BASEC 202300714) support these findings and emphasize the need for larger pediatric studies. 

Introduction: Although frequently reported by Multiple Sclerosis patients (MS), it is hard to objectify subtle changes in oculomotor function. An objective measure of oculomotor involvement might represent a biomarker alleviating lesion localization, assessing disease progression and response to treatment. We applied a novel eye-tracking tool (PRET™, machineMD AG, Switzerland)³ to objectively assess a quantitative neuroophthalmological status in moderately to severely disabled MS and healthy controls (HC).

Objectives: We aim to i) demonstrate feasibility of PRET™ in neurological conditions and ii) provide an objective measure of deviations in primary gaze and horizontal/vertical saccades.

Methods: MS and HC were recruited for a pilot study using PRET™, a virtual reality (VR-)headset-based oculography tool (ethics approval no. 23-7934-BR, Ruhr-University Bochum). We examined MS with an expanded disability status scale (EDSS) of ⩾3.0 and sex- and age-matched HC. PRET™ runs an automated programme examining gaze holding and horizontal and vertical saccades, among other parameters not evaluated, here.
Descriptive statistics include median [interquartile range, IQR] and absolute frequencies. Fisher’s exact and Mann-Whitney test were used for comparisons. Exact p-values are given.

Results: We examined 17 MS (47.1 [41.2-56.6] years; n=12 female, n=5 male; n=12 relapsing-remitting MS (RRMS), n=4 secondary progressive MS (SPMS), n=1 primary progressive MS (PPMS); EDSS 4.0 [3.0-5.5]) and 17 HC (47.4 [40.5-56.1] years; n=12 female, n=5 male). None of HC and 8/17 MS reported subjective ocular complaints. All participants underwent PRET™ examination without patient-sided abortion due to discomfort or motion sickness. Primary gaze abnormalities were detected in 7/16 MS and 2/16 HC (data missing in n=1 each, p=0.1134). Saccadic abnormalities were present in 10/17 MS and 3/17 HC (p=0.0324), mostly driven by horizontal saccadic abnormalities, such as subtle internuclear ophthalmoplegia (INO; 8/17 MS, 1/17 HC, p=0.0167; vertical: 5/17 MS, 3/17 HC, p=0.6880). In HC, presence of saccadic abnormalities was not associated with age (p=0.5088).

Conclusions: VR-examination was feasible and well-tolerated by all MS and HC. Primary gaze and saccadic abnormalities were detected in a higher proportion of MS with moderate to severe disability than HC with some specific patterns such as INO not detected during routine clinical examination. VR-based oculography may serve as an outcome tool if further confirmed during our study in different MS stages.

Background: Oculomotor symptoms are frequent in Myasthenia Gravis (MG) and significantly contribute to the burden of disease in affected patients. Nevertheless, reporting of performance-dependent occurrence of double vision can often not directly be reproduced during short-term clinical examination. Here, we applied a novel AI-supported tool (PRET™)³ to objectively assess a quantitative neuroophthalmological status in patients with severe generalized MG and to demonstrate feasibility of PRET™ in neurological conditions affecting the oculomotor system.

Methods: Patients with different neurological conditions and healthy controls are currently recruited for a prospective pilot study using PRET™, a novel virtual reality (VR-) headset-based oculography tool (ethics approval no. 23-7934-BR, 15-Nov-2023, ethics committee of the Medical Faculty of Ruhr-University Bochum). Within this framework, patients with severe generalized MG have been examined clinically and using PRET™ and scored (activities of daily living MG-ADL, quality of life MG-QoL). PRET™ runs an automated programme examining gaze holding and eye movements including horizontal and vertical saccades and horizontal smooth pursuit as well as ocular alignment, fusional amplitudes, visual fields and pupillary function (the latter not evaluated, here) within less than 12 minutes. We qualitatively summarize the data of the first three patients.

Results: We present the baseline data of three MG patients prior to their first infusion of an ongoing efgartigimod treatment, all of whom reporting end-of-dose symptom worsening before reinfusion (Pat. 1: male, 63 years; Pat. 2: male, 43 years; Pat. 3: female, 21 years). All patients reported double vision of varying degrees. They fully underwent PRET™ examination without patient-sided interruption or abortion due to discomfort or motion sickness.

As compared to three healthy participants with only few, patients 1 and 2 showed several saccadic intrusions during gaze holding, yet still being able to fixate with only minor intereye horizontal gaze position difference (below 2°). Patient 3 complaining constant double vision was not able to binocularly fixate in all positions except for right-up gaze (intereye horizontal gaze position difference ranging from 5 to 10° except for right-up gaze with 0-3°).

During smooth horizontal pursuit, performed approximately 7 minutes after start of the examination, the intereye difference is higher as compared to the initial phase of the examination in all three patients hampering binocular fixation. This phenomenon of exhaustion is not seen in either of the healthy participants.

Conclusions: The automated neuroophthalmological examination was feasible in all three patients and informed objectively about performance-dependent worsening of MG-associated oculomotor symptoms during the 12-minute examination. It clearly outperformed sole clinical examination and may serve as a novel outcome tool if further confirmed during our ongoing study.