Vertigo commonly arises from benign vestibular dysfunction but may also have a central cause such as stroke (in approximately 10% of cases) [1]. Nystagmus analysis is key to differentiating these disorders [2], yet conventional video-oculography (VOG) requires specialized laboratories and personnel, limiting access [3]. We developed a wearable augmented reality (AR)–based system delivering standardized oculomotor stimuli with real-time eye tracking. This study reports its design and preliminary validation.

One participant with prior cataract surgery could not calibrate the AR glasses, leaving 8 valid cases (mean age 60.4, SD 8.9 years; range 46-72). No significant discomfort or adverse effects were reported (Multimedia Appendix 4).

A total of 48 oculomotor data points were analyzed (saccades, pursuit, and gaze fixation). Agreement rates between AR and VOG ranged from 62.5% to 87.5% (Multimedia Appendix 5). Overall diagnostic accuracy was 77.1%, with sensitivity 81.8% (9/11; 95% CI 48.2%-97.7%), specificity 75.7% (28/37; 95% CI 58.8%-88.2%), PPV 50.0% (9/18; 95% CI 26.0-73.9%), and NPV 93.3% (28/30; 95% CI 77.9%-99.2%) (Multimedia Appendix 6). For central pathology, sensitivity reached 83.3% and specificity 100%. VAS scores did not differ significantly between AR and VOG, confirming tolerability.

This study demonstrates the feasibility of wearable AR glasses for nystagmus examination, showing diagnostic consistency comparable to VOG, particularly in ruling out central abnormalities. Results align with recent AR-based HINTS (head impulse, nystagmus, test of skew) assessments using head-mounted devices [6] and smartphone nystagmus apps with approximately 82% sensitivity [7].

Although enrollment spanned 4 months, only 8 participants completed paired assessments due to the 2-to-3-month delay for conventional VOG. This explains the small sample and illustrates the clinical bottleneck the AR system seeks to address. Patient tolerance was favorable, with no significant discomfort, consistent with prior AR studies [8]. One participant with cataract surgery could not be calibrated, likely due to altered ocular optics, underscoring the need for adaptive algorithms [9]. Unlike stationary VOG laboratories, wearable AR systems are portable and deployable in clinics, emergency care, or telemedicine, enabling point-of-care testing without specialized infrastructure [10]. With automated guidance, real-time tracking, and potential artificial intelligence integration, they may reduce reliance on experts and support decision-making.

Limitations include the small sample, yielding wide CIs for sensitivity (81.8%; 48.2%-97.7%) and specificity (75.7%; 58.8%-88.2%), reducing precision. The moderate PPV (50.0%; 26.0%-73.9%) highlights the need for improved processing. Diagnostic concordance was measured by percent agreement, which does not adjust for chance; future studies should apply Cohen κ and multiple raters. Finally, as a single-center pilot with one clinician, multicenter validation is required to confirm generalizability.