Sleep apnea screening still depends on a workflow that feels older than the problem it is trying to solve. Millions of patients with loud snoring, daytime sleepiness, resistant hypertension, or unexplained arrhythmias never make it to a sleep lab. Those who do often end up attached to belts, leads, cannulas, and finger sensors for a single night that is supposed to represent their normal sleep. The mismatch is obvious.
That is why camera-based sleep apnea screening keeps attracting serious research attention. A bedside camera can watch breathing effort, body motion, facial color changes, and recovery patterns all night without touching the patient. The pitch is not that video replaces polysomnography tomorrow. It is that contactless overnight monitoring could widen the funnel, catch more high-risk patients earlier, and make repeat screening practical outside the lab.
"An estimated 936 million adults aged 30-69 years worldwide have obstructive sleep apnoea, and 425 million have moderate to severe disease." — Benjafield et al., Lancet Respiratory Medicine (2019)
Why sleep apnea is a screening problem before it is a technology problem
The global burden is large, but the real issue is missed detection. Benjafield and colleagues' 2019 analysis in Lancet Respiratory Medicine put the worldwide number of adults with obstructive sleep apnea at 936 million. Review articles and sleep medicine groups still cite underdiagnosis rates in the 80% range in many populations. That means the bottleneck is not just treatment capacity. It starts much earlier, with who gets identified at all.
Polysomnography remains the reference standard because it measures airflow, thoracoabdominal effort, oxygen saturation, sleep stage, limb movement, and cardiac rhythm together. But PSG does not scale cheaply, and it does not travel well into ordinary bedrooms. Home sleep apnea tests reduce the burden, but they still require wearables or adhesive sensors, and those devices can fail or get removed in the middle of the night.
Camera systems are interesting for one simple reason: they shift the sensing burden away from the sleeper.
What overnight video systems actually measure
Most camera-based sleep apnea projects do not rely on a single signal. They combine several weak but useful clues.
Thoracoabdominal motion: Infrared or depth cameras can track chest and abdominal displacement over time. This matters because obstructive and central events produce different movement patterns.
Facial rPPG signals: Remote photoplethysmography extracts pulse-related color changes from exposed skin. Overnight, that can support heart rate tracking, respiratory modulation analysis, and recovery signatures after respiratory events.
Breathing-related facial cues: Near-infrared facial video can pick up subtle respiratory features even in dark conditions, especially around the nose, cheeks, and upper torso.
Body position and movement arousals: Sleep-disordered breathing is strongly shaped by position. A camera can often tell whether a patient is supine, lateral, restless, or undergoing frequent post-event movement.
Event morphology over time: The hardest part is not spotting a single pause. It is measuring repeated cycles of effort, cessation, recovery, and desaturation proxy signals across six to eight hours.
How camera-based sleep apnea screening compares with other approaches
| Method | Contact required | Main signals | Strength | Main limitation | Best role today |
|---|---|---|---|---|---|
| In-lab polysomnography | Yes | EEG, airflow, belts, SpO2, ECG, EMG | Full clinical reference | High cost and burden | Diagnosis |
| Home sleep apnea test | Yes | Airflow, effort, SpO2, pulse | More scalable than PSG | Sensor failure, fewer channels | Home diagnosis support |
| Wearable/PPG screening | Yes | Pulse, SpO2, movement | Consumer scale | Adherence and placement issues | Population screening |
| Bedside radar | No | Motion, respiration | Works in the dark | Dedicated hardware | Longitudinal home monitoring |
| Camera + infrared/rPPG | No | Motion, facial pulse, respiratory cues, position | Rich multimodal data from common optics | Sensitive to visibility, lighting, occlusion | Screening and overnight monitoring research |
The appeal of video is that it sits between PSG and simpler screening tools. It captures more context than a single pulse waveform, but it can be less burdensome than sensor-heavy testing.
Current research and evidence
One of the clearest demonstrations came from Sina Akbarian, Azadeh Yadollahi, and Babak Taati at the University Health Network and University of Toronto. In a 2020 JMIR study, their team analyzed overnight infrared video from 42 participants and used computer vision models to distinguish obstructive apnea from central apnea. Their best 3D-CNN model reached 95% accuracy with an F1 score of 89%. That result matters because separating obstructive from central events is clinically meaningful, and they did it from video-tracked chest and abdominal motion rather than attached effort belts.
A different line of work came from Clemens Veauthier and colleagues in Scientific Reports in 2019. Their group recorded 59 people simultaneously with polysomnography and a nocturnal 3-D camera system. The goal was not a full replacement for PSG. It was to test whether visual perceptive computing could estimate respiratory event burden in a way useful for home-style monitoring. That framing feels realistic. Most contactless systems will live or die on whether they can triage and trend well, not whether they can instantly replicate every PSG channel.
Researchers are also pushing facial-video methods in low-light settings. In 2023, Kurita and colleagues reported in Artificial Life and Robotics that near-infrared face video could recover respiratory-induced features correlated with reference respiratory measurements in the dark. That is important for bedroom deployment. A standard RGB pipeline can look impressive in daytime demos and fall apart once the lights go out.
There is also a broader signal-processing trend worth watching. Teams in sleep medicine and computer vision are moving toward multimodal models that fuse motion, facial optics, and temporal context instead of trying to solve apnea detection from a single handcrafted feature. That matches the clinical reality. Sleep apnea is a pattern problem.
Clinical applications under discussion
High-volume screening before formal testing
This is the most obvious use case. A contactless overnight camera system could flag patients with likely sleep-disordered breathing before they ever reach formal diagnostics. That is especially relevant in primary care, cardiology, and remote monitoring programs where symptoms are common but sleep testing is underused.
Monitoring treatment response over time
Patients do not just need a one-time label. They need to know whether positional therapy, oral appliances, weight loss, or PAP adherence changes nighttime breathing patterns. A contactless tool that can repeat measurements frequently has an advantage here.
Hospital and post-acute observation
Sleep-disordered breathing overlaps with opioid exposure, heart failure, neurologic disease, and post-surgical deterioration. Overnight camera monitoring in these settings could support respiratory surveillance without adding another wearable to already complex care.
The technical hurdles are real
Some of the hardest problems are mundane.
- Blankets can hide the torso and distort motion measurements.
- Side sleeping reduces face visibility and can break rPPG signal quality.
- Low light pushes systems toward infrared hardware, which adds cost and complexity.
- Multi-hour recordings create storage, battery, and privacy problems that short daytime demos do not reveal.
- Sleep labs have labeled events; ordinary bedrooms usually do not. That makes model development slower and messier.
There is also a validation problem. Papers often report strong performance on carefully curated datasets, but clinical buyers will want evidence across body types, skin tones, sleep positions, camera placements, and messy home environments. That bar is appropriate.
The future of camera-based sleep apnea screening
The near-term opportunity is not a magic bedroom camera that replaces sleep medicine. It is a better front door. Camera-based systems can make overnight respiratory screening more repeatable, less intrusive, and easier to deploy at population scale. If they work well, more patients with probable disease get routed into formal diagnosis instead of slipping through the cracks.
The longer-term story is broader than apnea alone. Once a camera can robustly follow breathing effort, pulse dynamics, recovery after events, and sleep-related movement, it starts to look less like a niche sensor and more like a contactless nighttime monitoring platform. That is where companies including Circadify see the field heading: multimodal overnight vital sign monitoring that begins with screening and expands as the evidence base catches up.
For now, the research is promising precisely because it is getting more specific. The best groups are no longer asking whether cameras can see something during sleep. They are asking which signals hold up all night, which event types can be separated reliably, and where contactless monitoring fits in the clinical workflow. Those are the right questions.
Frequently Asked Questions
Can a camera diagnose sleep apnea on its own?
Not today. Camera-based systems are best understood as screening or monitoring tools that can flag suspicious breathing patterns, respiratory pauses, and cardiorespiratory instability. Clinical diagnosis still depends on validated sleep studies and physician interpretation.
What signals can a camera capture during suspected sleep apnea?
Depending on the system, cameras can capture chest and abdominal motion, facial blood-volume pulse signals through rPPG, respiratory rate variability, body position changes, and in some setups oxygen-related trends or infrared breathing features.
Why are infrared cameras often used for overnight sleep monitoring?
Bedrooms are dark, and visible-light cameras lose signal quality overnight. Near-infrared illumination lets a system measure motion and physiological features without disturbing the sleeper with bright light.
How close is camera-based sleep apnea screening to routine clinical use?
The evidence base is growing, especially for respiratory event detection and contactless breathing analysis, but the field still needs larger real-world validation studies, standardized benchmarks, and clearer regulatory pathways before it becomes routine clinical infrastructure.
Related Articles
- Camera-Based Sleep Quality Assessment with rPPG — A broader look at overnight heart rate, HRV, respiratory rate, and sleep-related camera monitoring.
- Contactless Respiratory Rate Detection with rPPG Technology — Why respiratory sensing remains the backbone of contactless overnight monitoring.
- Contactless SpO2 Monitoring — How oxygen-related trends fit into the larger sleep-disordered breathing picture.