The Challenge of Diagnosing Vertigo

Vertigo, a sensation of dizziness or unsteadiness, is a common and often debilitating symptom that can stem from a variety of causes, ranging from benign peripheral vestibular disorders to more serious central nervous system conditions like stroke. Accurate diagnosis hinges on the meticulous analysis of nystagmus, involuntary rhythmic oscillations of the eyes. The presence, direction, and characteristics of nystagmus provide crucial clues to pinpointing the origin of the vestibular dysfunction.

Traditionally, standardized oculomotor assessment, including nystagmus testing, has relied on video-oculography (VOG). VOG systems, while effective, typically necessitate specialized laboratory equipment, controlled environments, and highly trained personnel. This inherent requirement for resources and expertise often creates significant barriers to access, particularly in outpatient clinics, emergency departments, and remote settings where telemedicine is increasingly employed. The logistical challenges associated with conventional VOG can lead to delays in diagnosis and treatment, potentially impacting patient outcomes, especially in cases where prompt intervention for central causes is critical.

Introducing Wearable AR for Vestibular Assessment

The research team sought to address these accessibility issues by developing and evaluating a wearable AR-based system designed to replicate the functionality of conventional VOG in a more portable and user-friendly format. The system comprises J7EF Gaze smart glasses, an Android-based portable device for processing, and a back-end platform for data management. The smart glasses are equipped with dual Si-OLED displays, providing a virtual visual stimulus, and a 30 Hz infrared eye-tracking sensor that captures eye movements with precision. An optional magnetic light shield can be attached to mimic the visual isolation provided by Frenzel goggles, a common tool used during nystagmus examination to prevent visual fixation from suppressing nystagmus.

The AR system utilizes Unity 3D software to generate a virtual display at a simulated distance of 1 meter. This display presents a series of standardized visual stimuli, including smooth pursuit eye movements along horizontal and vertical axes, fixation tasks, and saccades (rapid, ballistic eye movements). In-house software then delivers these stimuli and transmits real-time gaze data wirelessly to the back-end platform for automated analysis and secure storage. This integrated approach aims to standardize the testing procedure, minimizing variability that can arise from manual operation of conventional equipment.

Study Design and Methodology

To rigorously assess the performance of the AR system, the researchers implemented a randomized crossover usability study. This design is particularly well-suited for comparing two different interventions within the same participants, minimizing inter-individual variability. Nine patients diagnosed with vertigo were recruited for the study, which took place between October 2024 and January 2025 in a hospital-based clinical setting.

Each participant underwent two distinct testing sessions: one utilizing the wearable AR system and another employing conventional video-oculography. The order of these sessions was randomized to prevent any order effects. Crucially, a 30-minute washout period was incorporated between the two tests. This interval was designed to ensure that any residual effects from the first examination would not influence the results of the second.

Following the testing, the raw waveform outputs generated by both the AR system and the conventional VOG were pooled. To ensure objective evaluation, these pooled data were then independently interpreted by a board-certified otologist who was blinded to which system generated which data. This blinding is a critical step in preventing observer bias, ensuring that the interpretation of the results was solely based on the observed eye movements and not influenced by prior knowledge of the testing method.

Preliminary Findings: Feasibility, Usability, and Diagnostic Agreement

Of the nine participants initially enrolled, eight successfully completed both the AR-based and conventional VOG examinations. This high completion rate underscores the general feasibility of the AR system in a clinical setting. However, the study did identify a specific limitation: one participant with a history of cataract surgery experienced calibration difficulties with the AR glasses. This suggests that pre-existing ocular conditions affecting refractive properties might pose a challenge for the current AR calibration algorithms, highlighting an area for future technological refinement.

The analysis of oculomotor data revealed a notable degree of concordance between the AR system and conventional VOG. Across 48 distinct oculomotor data points, the agreement rates ranged from 62.5% to 87.5%. This variability in agreement across different oculomotor tasks warrants further investigation to identify which specific movements are most accurately captured by the AR system.

From a diagnostic perspective, the AR system demonstrated an overall diagnostic accuracy of 77.1%. The sensitivity, which measures the proportion of true positives (correctly identifying those with pathology), was 81.8%, while the specificity, which measures the proportion of true negatives (correctly identifying those without pathology), was 75.7%.

Of particular interest were the predictive values. The positive predictive value (PPV), the probability that a positive test result truly indicates the presence of disease, was 50.0%. A PPV of 50% indicates that when the AR system suggests a problem, there is only a 50% chance that it is a true positive. This moderate PPV suggests that while the AR system can flag potential issues, further confirmation might be necessary. Conversely, the negative predictive value (NPV), the probability that a negative test result truly indicates the absence of disease, was an impressive 93.3%. A high NPV is crucial for screening tools, as it indicates a low likelihood of missing actual cases. This suggests that the AR system could be particularly useful in ruling out certain vestibular pathologies.

When examining suspected central vestibular pathology, the AR system showed even stronger performance, with a sensitivity of 83.3% and a specificity of 100%. However, the researchers cautioned that these figures are based on an extremely small sample size and must be interpreted with significant reservation. Nevertheless, the perfect specificity for central vestibular pathology, even in this limited cohort, is an encouraging signal for its potential in differentiating serious conditions.

Patient Experience and Tolerability

Beyond technical performance, the usability and patient acceptance of the AR system were thoroughly evaluated. The study found that the AR system was well-tolerated by participants. Discomfort levels, measured using a visual analog scale, did not differ significantly between the AR testing and the conventional VOG. Importantly, no significant discomfort or adverse effects were reported by any of the participants during or after the AR examination. This favorable tolerability profile is a key advantage, especially for patients who may already be experiencing significant symptoms of vertigo.

Implications and Future Directions

The findings of this study carry significant implications for the future of vestibular diagnostics. The authors highlighted that wearable AR technology could play a vital role in overcoming the delays and access barriers currently associated with conventional vestibular testing. By offering a portable, potentially more accessible, and user-friendly alternative, AR systems could expand the reach of specialized vestibular assessments to a wider range of healthcare settings and patient populations. This could be particularly transformative for telemedicine, enabling remote consultations and diagnoses for individuals in underserved areas.

Despite the promising results, the researchers were candid about the study’s limitations. The very small sample size is a primary concern, leading to wide confidence intervals for the reported diagnostic metrics. The single-center design also limits the generalizability of the findings. Furthermore, the reliance on percent agreement rather than a more robust statistical measure like Cohen’s kappa for assessing diagnostic concordance and the moderate positive predictive value are areas that require attention in future research.

To address these limitations and pave the way for clinical implementation, the authors strongly recommended further studies. These should involve larger, multicenter cohorts to provide more statistically robust data and assess performance across diverse patient populations and clinical settings. The inclusion of multiple independent raters would also enhance the reliability of the diagnostic interpretations. Continued development of improved calibration algorithms, particularly to accommodate individuals with pre-existing ocular conditions, is also essential. Ultimately, a comprehensive, multicenter validation study is necessary before wearable AR systems can be confidently integrated into routine clinical practice for nystagmus examination in patients with vertigo.

The journey of integrating advanced technologies like AR into healthcare is often iterative. This initial study represents a significant step forward, demonstrating the potential of AR to democratize access to sophisticated diagnostic tools and improve the management of conditions like vertigo. As the technology matures and further validation is obtained, wearable AR systems could become an indispensable part of the diagnostic armamentarium, offering a more accessible, efficient, and patient-centered approach to vestibular health.

Citation

Wu CN, et al. Wearable augmented reality for nystagmus examination in patients with vertigo: randomized crossover usability study. J Med Internet Res. 2025;27:e75327. doi:10.2196/75327