What Our Smart Ring Sees at Night: Internal PPG Data vs PSG
3:17 AM. Our subject rolled over, and the PSG technician marked an arousal. Our ring? It caught the heart rate spike but missed the EEG change entirely.
I was sitting in the control room of our partner sleep lab, drinking terrible instant coffee and watching the polysomnography (PSG) waveforms scroll across the monitor. The electroencephalogram (EEG) showed a clear shift from N2 to a brief wakefulness, lasting exactly 14 seconds. Our smart ring, sitting snugly on the subject’s left index finger, registered a 12 bpm increase in heart rate and a slight spike in skin temperature, but its algorithm confidently kept the epoch scored as “Light Sleep.” We missed it completely.
That single night sparked a rigorous internal debate among our engineering team about the true limits of peripheral photoplethysmography (PPG). It also prompted us to conduct a massive, strictly controlled internal validation study. Today, I want to pull back the curtain on what our sensors actually see in the dark, where they excel, and where they still fall short compared to the clinical gold standard.
The Sensor Cocktail: How a Ring Actually Tracks Sleep
To understand our data, you first need to understand the hardware. A smart ring does not have the luxury of space that a chest strap or a bedside monitor enjoys. We are packing medical-grade capabilities into a 4-gram titanium shell.
At the core of our sleep tracking is the PPG sensor. We use a combination of green LEDs for continuous heart rate and heart rate variability (HRV) tracking, alongside red and infrared LEDs to measure blood oxygen saturation (SpO2). By analyzing the pulsatile changes in blood volume under the skin, we can extract a wealth of cardiovascular data. But PPG alone is blind to movement. That is why we pair it with a high-sensitivity 3-axis accelerometer. This acts as our actigraphy layer, detecting gross motor movements and micro-tremors.
The secret weapon in our ring, however, is the skin thermistor. Sleep onset is heavily correlated with vasodilation. As your core body temperature drops, your distal skin temperature rises. By measuring the distal-to-proximal temperature gradient at the finger, we can predict sleep onset latency with high accuracy. The PPG sensor in our ring uses the same signal processing pipeline we spent 15 years refining for medical watches — it’s why our resting heart rate RMSE is under 2 bpm.
The Gold Standard and Its Blind Spots
Polysomnography remains the undisputed gold standard for sleep assessment. When a patient is hooked up to a PSG, technicians are measuring brain waves (EEG), eye movements (EOG), and muscle tone (EMG), alongside respiratory effort and airflow. This allows them to definitively score sleep stages according to strict neurological criteria.
But PSG has massive ecological blind spots. When you strap a human being to a bed with two dozen electrodes, the data is undeniably pristine, but the natural sleep environment is destroyed. The bulky cap, the restrictive chest belts, and the noisy nasal cannula completely alter the very sleep architecture you are trying to measure. This frequently results in the well-documented “first-night effect,” where deep sleep is artificially suppressed and sleep fragmentation increases simply because the patient is uncomfortable.
A ring cannot measure brain waves. It cannot definitively see your eyes darting beneath your eyelids. But it can track your physiological state in your own bed, on your own mattress, for 365 nights a year without altering your natural behavior.
The Numbers: Our Internal 54-Night Trial
To bridge the gap between consumer convenience and clinical accuracy, we recently concluded an internal validation study. We recruited 18 subjects, ranging from healthy young adults to older individuals with mild sleep complaints. Over the course of the trial, we collected 54 nights of paired, synchronized recordings: our smart ring on the finger, and a full clinical PSG on the head.
We ran our raw PPG and accelerometer data through our proprietary algorithm and compared the output epoch-by-epoch against the PSG scored by certified registered polysomnographic technologists (RPSGTs).
| Metric | Smart Ring vs PSG Performance |
|---|---|
| Total Sleep Time (TST) Error | -8.2 minutes (Ring slightly underestimates) |
| Heart Rate RMSE (Resting) | 1.8 bpm |
| Heart Rate RMSE (With Movement) | 3.4 bpm |
| SpO2 Mean Bias | -0.3% (Ring slightly underestimates) |
| Sleep Stage Agreement: Wake (Cohen’s kappa) | 0.89 |
| Sleep Stage Agreement: Light Sleep (Cohen’s kappa) | 0.64 |
| Sleep Stage Agreement: Deep Sleep (Cohen’s kappa) | 0.58 |
| Sleep Stage Agreement: REM Sleep (Cohen’s kappa) | 0.52 |
Looking at the table, the cardiovascular metrics are incredibly tight. A resting heart rate RMSE of 1.8 bpm is well within the acceptable limits for continuous consumer monitoring. The SpO2 mean bias of -0.3% means our ring is highly compatible with clinical pulse oximetry for tracking general overnight trends, even if it isn’t calibrated for medical-grade apnea diagnostics.
Triumphs and Embarrassing False Positives
Where does the ring shine? Total Sleep Time and Wake detection. Our Cohen’s kappa for Wake is 0.89, which is exceptionally high for a wearable device. The combination of the temperature gradient dropping and the accelerometer detecting stillness makes wake detection highly reliable. Total sleep time error was only -8.2 minutes across 54 nights, meaning we slightly underestimate how long you sleep, likely because we are stricter about defining the exact moment of sleep onset.
But let’s talk about where we struggle. Differentiating Deep Sleep from REM Sleep is the Achilles heel of all peripheral wearables. Our kappa for Deep is 0.58, and for REM, it drops to 0.52. Why? Because both stages feature profound physical stillness. The only way to tell them apart without an EEG is through heart rate variability. REM sleep features erratic, sympathetic-driven heart rate spikes, while Deep sleep features slow, parasympathetic-driven bradycardia.
I still cringe thinking about Subject 04 during our pilot phase. He was a restless sleeper who liked to read in bed with a heavy book propped on his chest. Because his arm was perfectly still and his heart rate dropped into a relaxed baseline, our algorithm confidently scored 45 minutes of “Deep Sleep” before he actually closed his eyes. We had to manually discard that epoch. It was a humbling reminder that peripheral perfusion and stillness do not always equal unconsciousness. If the finger curls slightly and the PPG signal degrades, the algorithm loses the HRV nuances required to separate REM from Deep sleep.
Clinical Implications: Consumer Wellness vs. Diagnosis
This brings us to a critical distinction in the wearable health space. What is our data actually good for?
The AASM Manual for the Scoring of Sleep and Associated Events strictly requires EEG, EOG, and EMG for definitive sleep staging and the diagnosis of primary sleep disorders. A smart ring cannot support a clinical diagnosis of narcolepsy, periodic limb movement disorder, or severe obstructive sleep apnea. If you suspect you have a clinical sleep disorder, you need an attended PSG or a certified Type III home sleep test.
However, for longitudinal consumer wellness tracking, the utility is immense. As highlighted in a 2022 comparative study in Sleep Medicine evaluating wrist-worn PPG devices, while consumer wearables struggle with epoch-by-epoch clinical staging, they are highly accurate at tracking macro-level sleep trends, circadian rhythm disruptions, and recovery metrics over time. Our ring is fully compatible with these wellness use cases. It is designed to tell you how your alcohol intake affected your deep sleep last night, or how your training load is impacting your resting heart rate over a 30-day macrocycle. It is a tool for behavioral modification, not a diagnostic medical device.
Let’s Build the Next Generation Together
At Geyan Technology Innovation, we believe in radical transparency about what our hardware can and cannot do. We are constantly refining our signal processing pipelines to push the boundaries of what a 4-gram ring can achieve, but we know that peripheral PPG will always have physiological limits compared to a full EEG setup.
We are currently opening our OEM and ODM channels for the next iteration of our smart ring platform. If you are a brand looking to integrate clinical-grade PPG signal processing, advanced sleep staging algorithms, and robust hardware design into your own product line, we want to talk.
Reach out to our partnerships team to discuss technical specifications, integration capabilities, and volume manufacturing.
Contact: Lin Jie, CEO
Email: jine@xdunmedical.com
Phone/WhatsApp: +86-13544254314
Website: xdunmedical.com