Can a Smartphone Detect Nystagmus? What the Research Shows
Can a smartphone detect nystagmus? An honest review of the 2026 UCL emergency-room study and other phone eye-tracking research, and what is still unproven.
EyeRehab - VOR Training Team
Published on November 20, 2026
Listen to this article
Natural Language Narration · 17 min
Quick answer
Can a smartphone detect nystagmus?
Not reliably on its own yet. Research shows an ordinary smartphone can record eye movements well enough for a clinician to review, and a 2026 University College London study found handheld emergency-room videos carried useful eye-movement signals. Automated detection looks promising in small studies but is not validated for unsupervised use, so no app should be used to rule out a stroke.
Reviewed on November 20, 2026
Can a smartphone detect nystagmus?
A smartphone can record nystagmus well enough for a clinician to see it, and in small studies phone software has measured or flagged it under controlled conditions. No app has yet been validated to pick out nystagmus on its own in real patients without supervision, so today a phone video is something to share with a clinician, not a test that rules anything out.
Nystagmus is an involuntary, rhythmic movement of the eyes, usually a slow drift followed by a quick corrective jerk. The reference tools for measuring it are videonystagmography (VNG) and video-oculography (VOG), in which infrared cameras in goggles track the pupils. The video head impulse test (vHIT) uses similar goggles to measure how well the eyes stay on target during quick head turns. For the basics, see What Is Nystagmus?
Phones appeal because nystagmus often stops before a person reaches a specialist, goggles are scarce, and the eye signs are hard to read. A 2024 study in the Journal of the American Heart Association (Bastani and colleagues) noted about 5 million dizziness visits to US emergency departments each year, more than 250,000 of them vestibular strokes, and named unfamiliarity with the subtle eye signs as a major barrier.
It helps to separate three jobs:
| Job | What it means | State of the evidence |
|---|---|---|
| Record | Capture a video a clinician can watch | Strongest: real emergency-room videos and self-recorded tests were usable |
| Measure | Turn the video into an eye-position trace with speed and direction | Promising: close agreement with lab goggles in small, controlled studies |
| Decide | Software flags nystagmus by itself, with no human review | Early: tiny samples, few or no negative cases, not validated for unsupervised use |
What did the 2026 UCL emergency-room study find?
It found that ordinary handheld phone videos from a busy emergency department carried enough eye-movement information to track nystagmus, even beats too small to see easily. It did not test telling nystagmus from no nystagmus.
The study, “Real-world handheld smartphone videos for nystagmus detection in acute vertigo: a feasibility study,” was published in the Journal of Neurology in 2026 by Cortese, Duvieusart, Khan, Leung, Koohi and Kaski, a team at University College London and UCL Queen Square. The videos came from a UK tertiary-hospital emergency department.
How the study worked
- Videos: 67 short handheld clips, 3 to 5 seconds long, from the SMART-VERTIGO dataset. All came from patients with acute vertigo in whom clinicians had identified nystagmus.
- Phone: A mid-range Android phone (Samsung Galaxy A55), held by hand with no tripod or stabilizer, recording at 30 or 60 frames per second.
- Tracking: The method followed keypoints on the eye and on the surrounding skin. The skin moves with camera shake but not with the eye, so the shake could be subtracted. It extracted horizontal eye-movement traces, and nystagmus episodes were identified by signal analysis checked against the video.
What it found
- Tracking was possible in 63 of the 67 videos (94%), and 40 met the criteria for nystagmus analysis.
- Nystagmus seen on video was reflected in the traces, including very low-amplitude nystagmus barely visible to the naked eye.
- This held despite poor lighting, shadows, camera motion and eye make-up.
What it did not show
- Blinking was the main reason tracking failed.
- Manual steps: some keypoint selection and the segmentation were done by hand.
- No negative cases: every video showed nystagmus, so the study did not test telling nystagmus apart from no nystagmus.
- No VNG reference: without goggle recordings to compare against, accuracy could not be measured.
- Horizontal only: vertical and twisting (torsional) movements were not analyzed.
- Frame rate: at 30 or 60 frames per second, a phone cannot fully capture very fast eye movements.
The authors concluded that handheld, low-cost smartphones can capture clinically meaningful eye-movement information in real emergency care, laying the groundwork for scalable, telehealth-compatible tools. They wrote that further studies are needed to automate detection and to validate performance in larger prospective cohorts.
Eye Rehab’s eye clips were inspired by this study. We built on its finding that an ordinary handheld phone video can carry useful eye-movement information. The study did not involve or test Eye Rehab.
How accurate are smartphone eye-tracking apps compared with VNG?
In small studies, phone tracking agreed closely with lab goggles when nystagmus was induced in healthy volunteers. Evidence in real patients is much thinner.
| Study | What the phone did | Setting and people | Main limit |
|---|---|---|---|
| Cortese et al., J Neurol 2026 | Tracked horizontal eye movement in handheld videos | 67 emergency-department videos, acute vertigo with nystagmus | No videos without nystagmus, no VNG comparison, some manual steps |
| Friedrich et al. (ConVNG), J Neurol 2023 | Neural network tracked the pupil to measure induced nystagmus | 10 subjects, compared with VOG | Small; induced nystagmus; precision below VOG |
| van Bonn et al., Eur Arch Otorhinolaryngol 2022 | Custom app tracked induced nystagmus | 20 healthy young adults | Healthy volunteers only; abstract reports sensitivity only |
| Bastani et al. (EyePhone), JAHA 2024 | Measured the speed of induced nystagmus | 10 healthy volunteers, compared with VOG | Healthy volunteers; induced nystagmus |
| Bastani et al., Digit Biomark 2025 | Let people film their own positional tests | Healthy volunteers | Usability only; no one had BPPV |
| Phillips et al., Digit Biomark 2025 | Flagged positional nystagmus from eye traces | 23 videos from 10 patients with suspected BPPV | Only 3 positive videos |
| Parker et al. (eyePhone), Digit Biomark 2021 | Measured the video head impulse test | 11 clinic patients, compared with VOG vHIT | Proof of concept; head impulses, not nystagmus |
| Duvieusart et al., IEEE EMBC 2025 | Deep-learning pupil tracking | 12 induced-nystagmus videos, compared with a commercial goggle system | Small test set; moderate agreement |
Induced nystagmus. Several studies used optokinetic nystagmus, the jerking eye movement anyone makes while watching a moving striped pattern, because it can be produced on demand. ConVNG (Friedrich and colleagues, 2023) fine-tuned a convolutional neural network on more than 550 annotated frames to track the pupil in smartphone video. In 10 subjects, its slow-phase velocity measurements (the speed of the slow drift) were statistically equivalent to VOG. Its precision of 0.30 degrees per second beat the general-purpose MediaPipe tool (0.7) but not VOG (0.12). In the EyePhone study (Bastani and colleagues, 2024), 10 healthy volunteers watched stimuli moving at 2 to 12 degrees per second. Phone measurements correlated closely with VOG (r = 0.98 horizontal, 0.94 vertical), and the area under the curve for picking out nystagmus was 0.87. Duvieusart and colleagues (2025) tested deep-learning pupil tracking on 12 optokinetic videos. Against the ICS Impulse goggle system, it captured the nystagmus “reasonably well” (average R² about 0.76), with refinement still needed.
van Bonn and colleagues (2022), in Rostock, Germany, built an app meant to help separate stroke from inner-ear vertigo. In 20 healthy volunteers with nystagmus induced by caloric (ear-warming or ear-cooling) and optokinetic stimulation, sensitivity averaged 82% in the optokinetic trials. The stroke question itself was not tested.
Positional nystagmus. In a 2025 usability study (Bastani and colleagues), healthy volunteers used EyePhone to film their own Dix-Hallpike and supine roll tests, the positional tests for BPPV (benign paroxysmal positional vertigo). All recorded their eye movements, and experts rated 95% of the videos as having clear eye visibility. A 2025 pilot (Phillips and colleagues) recorded 23 positional-test videos from 10 patients with suspected BPPV at Johns Hopkins Hospital. Experts found positional nystagmus in 3. The app’s traces flagged all 3 and ruled it out in the other 20 (sensitivity 100%, 95% confidence interval 44 to 100%; specificity 100%, 84 to 100%). The wide intervals reflect how few positives there were.
Head impulses. HINTS is a bedside exam for constant vertigo (Head Impulse, Nystagmus, Test of Skew) that trained clinicians use to separate stroke from inner-ear causes. Parker and colleagues (2021) tested the head impulse part with a phone app in 11 patients at an otoneurology clinic. Reflex gains from the phone and from VOG-based vHIT correlated highly (r = 0.9) across 20 ears, and one patient’s test could not be quantified with either device.
Why is automatic nystagmus detection so hard?
Real-world video is messy, nystagmus can be tiny or twisting, and the studies so far are too small, with too few negative cases, to show how often an app would be wrong.
- Blinks and closed eyes: Blinking was the main cause of failure in the UCL study, and dizzy people often squint or shut their eyes.
- Light, shadow and make-up: The UCL tracking coped, but an automatic verdict must cope every time.
- Frame rate: Phones usually record at 30 or 60 frames per second, which limits measurement of fast eye movements.
- Direction: Most phone methods measure horizontal movement best. The twisting nystagmus typical of BPPV is hard to measure from a flat image.
- Fixation: Looking at a screen or a dot gives the eyes a target, which can damp inner-ear nystagmus. A quiet trace does not mean no nystagmus.
- Missing negatives and small samples: The UCL study had no videos without nystagmus and the Phillips pilot had only 3 positive videos, so neither can estimate false alarms or misses in a large, mixed group.
- Healthy volunteers: Induced nystagmus is cleaner than the nystagmus of stroke, neuritis or BPPV.
Validation for unsupervised use would need large prospective studies in real settings, including people with and without nystagmus and with stroke, against VNG and the final clinical diagnosis. A false all-clear during a stroke could be dangerous.
Can an app rule out a stroke?
No. No smartphone app has been validated to rule out a stroke in someone with dizziness, and you should never wait for an app or a video before calling for help.
For patients, the message is: record and share, but don’t rely on any app to tell you you’re safe. A short video from a spell shows your clinician what they missed; How to Record Nystagmus With Your Phone explains how to film safely. Don’t try to read your own video, because some strokes cause no nystagmus at all, as Nystagmus and Vertigo: What Your Eyes Tell Your Doctor explains. If an app promises a verdict, ask which study supports it and whether it included people without nystagmus.
How can clinicians use patient-recorded eye videos?
As a patient-recorded observation that adds to the history and exam, not as a measurement.
- Episodic disorders: BPPV, Ménière’s disease and vestibular migraine often look quiet in clinic. A clip from the spell can guide which tests to prioritize.
- Triage and telehealth: A clip can help decide how soon someone needs an in-person visit.
- Known limits: Frame rate, fixation, no torsion measurement and unknown distance and lighting. Use slow motion and frame stepping, and note what you could not see.
- Honest documentation: Record it as patient-recorded video, separate from HINTS by a trained examiner, VNG or vHIT.
For workflow details, see nystagmus eye clips for clinicians, Eye Rehab for physical therapists and our guide to VNG and balance testing in the vestibular clinic.
Where does Eye Rehab stand?
We chose capture plus clinician review, with no automatic verdict, because that is what the evidence supports today.
In the free Eye Rehab app (on Android now, iPhone support coming), a patient logs a dizzy spell, taps Record your eyes and looks at a dot for 3 to 10 seconds. The app records video only, no sound, and a helper can film with the back camera. The phone crops the video to the brows, eyes and bridge of the nose before anything leaves it and deletes the full-frame recording. The clip is stored encrypted with the episode, and sharing is a separate switch for each clinician, off by default.
In Eye Rehab Pro (no extra plan needed), a clinician the patient has shared with sees an Eye clips block labeled “recorded by the patient · not a measurement.” Playback offers frame-by-frame stepping, 0.25×, 0.5× and 1× speed and looping, with the camera used, who filmed, the framing check, length and frame rate. A “What did you see?” note (nystagmus, no nystagmus or can’t tell, plus beat direction) stays private to that clinician and feeds no score, rule or recommendation.
The app does not analyze the video, measure eye movement or tell anyone what the eyes are doing. It is not an emergency service and does not replace an examination, VNG or video head impulse testing. Patients can learn more about Eye Rehab eye clips.
When to get emergency help
Call emergency services right away if dizziness comes on suddenly with a drooping face, arm or leg weakness or numbness, trouble speaking, double vision, a sudden severe headache, being unable to walk or stand, or new hearing loss with vertigo. Also go to the emergency department the same day for new, constant, severe vertigo that lasts hours, even without these signs. Filming must never delay calling for help. Never film while driving or standing unsteadily; sit or lie down somewhere safe first. Is Nystagmus Dangerous? lists more warning signs.
Key Takeaways
- Recording works: Ordinary handheld phone videos can carry useful eye-movement information, even in an emergency department.
- Measuring is promising: Phone tracking has agreed closely with lab goggles in small, controlled studies.
- Automatic verdicts are not ready: Samples are small, negative cases are few or missing, and no app is validated for unsupervised use.
- No app rules out a stroke: Red flags mean calling emergency services, not filming.
- Eye Rehab captures, clinicians decide: Eye clips are patient-recorded and shared by choice, with no automated analysis.
Record and Share With Eye Rehab
Eye Rehab is free for patients. On Android, patients can record short eye clips during dizzy spells and share them with the clinician they choose. Eye clips are coming to the iPhone app. Looking for care? Find a vestibular clinician near you.
Medical Disclaimer
This content is for informational and educational purposes only and does not constitute medical advice, diagnosis or treatment. The studies described are early research, and none validates a smartphone app for unsupervised use. Eye Rehab eye clips are patient-recorded videos, not measurements, and only a clinician can interpret them. If dizziness comes with any stroke warning sign listed above, call emergency services immediately.
Related reading
- What Is Nystagmus? Types, Causes and When to Get Help
- How to Record Nystagmus With Your Phone for Your Doctor
- Nystagmus and Vertigo: What Your Eyes Tell Your Doctor
- Is Nystagmus Dangerous? Warning Signs and When to Get Help
- VNG and Balance Testing in the Vestibular Clinic
- Is It Your Eyes or Your Inner Ear? Dizziness Explained
Frequently Asked Questions
Can a smartphone detect nystagmus?
A smartphone can record nystagmus well enough for a clinician to see it, and small studies show phone software can measure or flag it under controlled conditions. But no app has been validated to pick out nystagmus on its own in real patients without supervision. Today, a phone video is something to share with a clinician, not a test that rules anything out.
What did the 2026 UCL emergency-room study find?
Researchers at University College London analyzed 67 handheld smartphone videos filmed in a UK emergency department from people with acute vertigo and nystagmus. Eye movements could be tracked in 63 of them (94%), and the nystagmus seen on video showed up in the traces, even very small beats, despite poor lighting and camera shake. The study had no videos without nystagmus and no VNG comparison, so it did not measure diagnostic accuracy.
How accurate are smartphone eye-tracking apps compared with VNG?
In small studies of healthy volunteers watching moving stripes, phone tracking agreed closely with lab goggles. One app's eye-speed measurements correlated with video-oculography at r = 0.98 horizontally and 0.94 vertically. Precision was still lower than lab equipment, and studies in real patients are much smaller, so phones are not yet a substitute for VNG.
Can an app rule out a stroke?
No. No smartphone app has been validated to rule out a stroke in someone with dizziness. Some strokes cause no nystagmus at all, and phone videos have limits in frame rate, lighting and fixation. If sudden dizziness comes with a drooping face, weakness, speech trouble, double vision, a severe headache, new hearing loss or being unable to walk, call emergency services instead of filming.
How can clinicians use patient-recorded eye videos?
As a patient-recorded observation that adds to the history and exam. A clip from a spell can help with triage, telehealth follow-up and episodic conditions that look quiet at the visit. Clinicians should review it with slow motion and frame stepping, note what they could and could not see, and keep it separate from measured findings such as HINTS by a trained examiner, VNG or video head impulse testing.
Tags
Written by
EyeRehab - VOR Training Team
Expert insights on vestibular rehabilitation and eye health.
Related Articles
VNG and Balance Testing in the Vestibular Clinic: A Practical Diagnostic Workflow for Clinicians
How videonystagmography and balance testing fit into a clinician's vestibular diagnostic workflow, what the current evidence supports, and how to move from test results to a rehabilitation pathway.
Eye Rehab Editorial Team
November 9, 2026
Neuroplasticity and Vestibular Rehabilitation: Hope for Recovery
Discover how brain plasticity enables vestibular recovery, offering new insights into healing neural pathways after injury or disorder.
VOR Eye Rehab Team
January 26, 2026
Dual-Tasking: Vestibular Training for Cognitive Health
Explore how vestibular rehabilitation can improve cognitive function, balance, and brain processing through integrated exercise strategies.
VOR Eye Rehab Team
December 31, 2025