Sleep is one of the strongest indicators of overall health, yet understanding the risks associated with poor sleep has traditionally required detailed observation and clinical expertise. Traditional diagnostics remain essential; however, emerging technologies are making early detection faster, simpler, and more accessible.
At Soliish, we’re working towards that goal.
AI-powered facial analysis represents a new dimension in sleep science where human expertise meets intelligent automation. By combining facial biometrics with machine learning, Soliish uses a simple image as an enabler for advanced sleep-health assessment—helping clinicians diagnose more intelligently, efficiently, and accurately.
Facial features often provide subtle clues about sleep quality — signs of both healthy rest and sleep disruptiion. Over the last decade, researchers have shown that sleep deprivation and sleep apnea can visibly alter facial features, and that these changes can partially or fully reverse following appropriate sleep therapy.
In one of the studies on this topic, Chervin et al. (2013) used high-precision 3D facial photography to capture people with obstructive sleep apnea (OSA) before and after two months of continuous positive airway pressure (CPAP) therapy.
Twenty-two independent raters, blinded to which photos were taken before or after treatment, consistently judged the post-treatment faces as more alert, youthful, and attractive. The researchers also measured real changes: reduced forehead puffiness and less redness around the cheeks and eyes, indicating improved circulation and reduced inflammation.
A few years later, Eastwood et al. (2020) demonstrated that these visible differences could even be detected by computers. Using 3D facial scans and machine learning, their model identified adults with sleep apnea with about 91% accuracy significantly outperforming traditional questionnaires. The machine learning model didn’t rely on tired expressions; instead, it recognized subtle structural differences in the face, including increased width, facial depth, and neck size, that correlate with airway obstruction.
More recently, Prikladnicki et al. (2024) showed that even trained human observers can spot tell-tale signs. Their team developed the CASA score (Cheeks Appearance for Sleep Apnea), rating cheek fullness and looseness from simple photos. The score alone predicted sleep apnea with around 86% accuracy nearly matching more complex screening tools.
Together, these studies reveal that our faces mirror what’s happening during sleep.
Chronic poor sleep or untreated apnea doesn’t just affect rest - it’s often reflected in facial structure. Subtle anatomical traits around the jaw, palate, and midface can signal an increased risk of airway collapse during sleep.
These findings paved the way for the modern AI system like Soliish’s. They turn what clinicians have long been observed; the facial indicators of sleep health into measurable data.
Artificial Intelligence is redefining how we interpret the human face. It is no longer a static image, but a dynamic map of physiological signals. Using advanced algorithms, AI systems process high-resolution facial images to identify minute changes linked to sleep and breathing.
What makes AI facial analysis remarkable is its non-invasive and instantaneous nature. This early-screening capability bridges preventive care with precision diagnostics.
It acts as an early line of observation identifying when deeper diagnostic assessments may be required.
We didn't reinvent sleep science. We just found a faster way to apply what we already know.
Step 1: Capture a photo
Any decent smartphone camera works. No special lighting, no makeup removal required. The image needs to be clear and front-facing.
Step 2: AI extracts facial features
The system evaluates key craniofacial features captured in the selfie, for example jaw alignment, midface dimensions, and facial proportions.
Step 3: Pattern matching against sleep data
It uses advanced algorithms trained on validated clinical datasets. These measurements are processed within seconds to identify patterns linked to airway restriction and sleep-apnea risk.
We've trained our models on facial images paired with real sleep test results (clinically validated polysomnography data). The AI model looks for patterns that correlate with sleep disorders.
Step 4: Generate a risk score
The system then provides a clear output—low, moderate, or high risk for sleep issues—along with the specific indicators that informed the classification. It also flags cases where image quality is too poor for reliable analysis.In just a few steps, Soliish turns a simple photo into meaningful sleep insights - bridging anatomy, AI, and accessibility for smarter screening.
Soliish’s technology makes understanding sleep health simpler, faster, and more accessible for everyone. Each advantage is designed to support real-world care; whether you’re in a clinic, at home, or working remotely.
1. Speed: Real-time analysis delivers immediate insights.
2. Objectivity: AI-driven evaluation eliminates subjective bias and human fatigue.
3. Scalability: Designed to handle high volumes across clinics, wellness centers, and research programs.
4. Remote Accessibility: Ideal for telehealth environments, enabling sleep assessments beyond the clinic.
Together, these benefits help clinicians and partners act sooner, personalize care, and bring smarter sleep health solutions to more people.
Soliish has been developed to complement existing sleep diagnostic systems. Its flexible architecture allows easy integration with digital health platforms, wearable data, and electronic health records, ensuring a unified view of patient sleep health.
By merging the accuracy of AI with the intuition of medical expertise, Soliish empowers professionals to act earlier, respond faster, and bring sleep health within everyone’s reach - one image at a time.
Find answers to frequently asked questions about our technology and services.
Research shows that sleep apnea–related facial traits often include increased facial width, midface depth, jaw alignment differences, cheek puffiness, or looseness. These subtle anatomical markers can signal airway restriction or collapse during sleep.
AI analyzes craniofacial structures—such as jaw position, midface proportions, and facial depth - using machine learning models trained on real sleep-study data (polysomnography). It identifies patterns that correlate with airway obstruction and sleep apnea risk.
Untreated sleep apnea can cause visible changes like puffiness, eye redness, facial swelling, and subtle shifts in facial structure over time. Studies show that after CPAP therapy, many of these effects improve due to better oxygenation, reduced inflammation, and more consistent sleep.
Jaw structure - such as a recessed chin, narrow jaw, or misalignment—can reduce airway space during sleep. Soliish’s facial scan flags craniofacial patterns associated with airway restriction, so you know when a professional evaluation may be needed.
A clear, front-facing selfie taken in normal lighting is enough. No special equipment, makeup removal, or clinical setup is required.
Yes. The analysis uses only surface-level facial images. There is no physical contact, radiation, or sensor-based measurement.
Subscribe to our newsletter to get the latest sleep health news, tips, and updates from Soliish delivered to your inbox
