Circadify
Roughly 13.4 million babies are born preterm each year worldwide. About 3 to 10 percent of all neonates end up in a neonatal intensive care unit, where continuous monitoring is standard — heart rate, respiratory rate, oxygen saturation, temperature, all tracked around the clock through sensors stuck to skin that bruises if you look at it wrong.
That last part is the problem nobody outside neonatology talks about. Preterm neonates have skin so thin and fragile that the adhesive electrodes used by conventional bedside monitors routinely cause injury. A 2023 quality improvement study in BMJ Open Quality found that adhesive-related skin injury rates in NICUs range from 9.25% to 41.5% of patients. The injuries concentrate on the face, arms, hands, and chest — exactly where sensors get placed. Oximeter probes, which must be tightly attached for accurate readings, can cause pressure sores or burns when not repositioned frequently enough.
There's a less obvious cost too. The tangle of wires restricts infant positioning, gets in the way of routine nursing care, and puts a physical barrier between parents and their baby. Kangaroo care — skin-to-skin contact that research has repeatedly shown benefits preterm outcomes — becomes logistically difficult when the baby is tethered to a monitor by multiple leads.
"Current methods to acquire vital signs are challenging to patients, parents, and health care professionals, as vital signs are usually obtained by the use of skin sensors connected to bedside monitors via wires." — Williams et al., Pediatric Research (2025)
What camera-based monitoring actually looks like in a NICU
The setup is simpler than you'd expect: mount a camera above the incubator or cot, capture video of the infant, and use algorithms to pull physiological signals out of the footage. Heart rate comes from detecting tiny color changes in the skin caused by blood pulsing through capillaries — the same remote photoplethysmography (rPPG) principle used in adult monitoring, but applied to a much smaller and more fragile subject. Respiratory rate comes from tracking the subtle chest and abdominal movements visible on video.
In practice, neonatal monitoring presents challenges that adult rPPG doesn't face. The signal is weaker because neonates are smaller. Skin pigmentation affects signal quality differently in infants than in adults. NICU lighting varies between units and even within a single unit across shifts. Infants move unpredictably — and unlike adults, you can't ask them to hold still for 30 seconds.
Estévez et al. published a 2025 study in Scientific Reports testing RGB-D cameras (standard color plus depth sensing) for continuous non-contact vital sign monitoring of neonates. Their system measured heart rate, respiratory rate, and oxygen saturation without any physical contact. The depth channel helped with motion compensation — a meaningful addition, since infant movement is one of the primary sources of measurement noise.
Comparing neonatal vital sign monitoring technologies
| Technology | Contact Required | Skin Injury Risk | Vital Signs Measured | Accuracy (HR) | Accuracy (RR) | Kangaroo Care Compatible | Clinical Readiness |
|---|---|---|---|---|---|---|---|
| Standard wired monitors | Yes — adhesive electrodes | High (9-42% injury rate) | HR, RR, SpO2, Temp | Gold standard | Gold standard | Limited — wires restrict | Clinical standard |
| Wireless wearable patches | Yes — adhesive or band | Moderate — reduced adhesive | HR, RR, Temp, SpO2 | ±2-3 bpm | ±2-4 brpm | Improved — no wires | Some clinical trials |
| RGB camera (rPPG) | None | None | HR, RR, HRV, RRV | ±2-5 bpm (controlled) | ±3-6 brpm | Fully compatible | Research stage |
| RGB-D camera (depth + color) | None | None | HR, RR, SpO2 estimate | ±2-4 bpm (controlled) | ±2-5 brpm | Fully compatible | Research stage |
| Thermal imaging | None | None | RR, Temp estimate | N/A — indirect only | ±2-4 brpm | Fully compatible | Research stage |
| Radar-based (FMCW) | None | None | HR, RR | ±3-5 bpm | ±2-5 brpm | Fully compatible | Early research |
Sources: Zeng et al. (2024), Estévez et al. (2025), Williams et al. (2025), Nagy et al. (2021).
The trade-off is obvious from the table: contact-based systems are accurate and clinically validated, but they hurt fragile skin and get in the way of parent-infant bonding. Non-contact systems solve those problems but can't yet match the reliability that standalone clinical decision-making requires.
Key research and who's doing it
Zeng, Yu, and Wang (2024) at Eindhoven University of Technology published what may be the most comprehensive camera-based neonatal study to date in IEEE Transactions on Instrumentation and Measurement. Their team collected video data from 50 preterm infants across two NICUs over two years. They developed a framework that extracts heart rate, respiratory rate, heart rate variability, respiratory rate variability, and actigraphy — all from a single camera feed. The dual-center design is significant because most prior studies operated within a single hospital, limiting generalizability. Their system tracked individual physiological patterns over time and could model group-level health trajectories for the cohort.
Estévez et al. (2025) at the University of Bristol used RGB-D cameras to monitor neonatal vital signs continuously, publishing their results in Nature Scientific Reports. Their approach combined the standard rPPG color-channel analysis with depth information to improve motion robustness. The depth data helped distinguish genuine physiological signals from movement artifacts — a persistent challenge in neonatal monitoring where infants shift position frequently.
Nagy et al. (2021) developed algorithms specifically for continuous camera-based monitoring of premature infants, publishing in Applied Sciences. Their system could measure pulse rate and breathing rate while also recognizing situations like medical intervention or high infant activity that would otherwise corrupt the signal. That context-awareness matters: in a real NICU, nurses are constantly interacting with patients, and a monitoring system that can't distinguish a nurse repositioning an infant from a genuine vital sign change will generate useless false alarms.
Villarroel et al. (2019) at Oxford tested non-contact vital sign monitoring on 30 neonates in a clinical setting, comparing camera-derived heart rate and respiratory rate against reference monitors. Their results, published in the British Journal of Anaesthesia's conference proceedings, showed strong correlation for heart rate and reasonable performance for respiratory rate, with the expected degradation during movement.
Williams et al. (2025) conducted a systematic review in Pediatric Research covering the entire landscape of next-generation NICU monitoring — both non-contact and wireless wearable approaches. Their review mapped out where each technology stands and concluded that while several systems show real promise, none have crossed the validation threshold needed for regulatory clearance as standalone monitors.
13.4M
Preterm Births Annually Worldwide
9-42%
NICU Skin Injury Rate From Adhesives
50
Infants in Largest Camera Monitoring Study
Clinical applications taking shape
Continuous monitoring without skin contact
The most realistic near-term application is supplemental monitoring. A camera system running alongside conventional monitors could provide a redundant vital sign stream without adding any more adhesives to a neonate's skin. If the camera reading diverges significantly from the contact sensor, it flags for attention. If the contact sensor fails or needs removal for skin recovery, the camera provides continuity. Several research groups have framed their work this way — not as a replacement, but as an additional safety layer that happens to cost nothing in terms of patient contact.
Supporting kangaroo care
Kangaroo care — extended skin-to-skin contact between parent and infant — has strong evidence behind it for improving preterm outcomes, including reduced mortality, fewer infections, and better neurodevelopmental results. But it's hard to do when the baby is wired to a monitor. Parents report anxiety about dislodging sensors, and nurses sometimes delay or shorten kangaroo care sessions because of monitoring logistics. A camera that continues tracking heart rate and breathing during skin-to-skin contact could remove that barrier entirely. The infant stays monitored, the wires come off, and the parent-infant bond isn't interrupted by equipment.
Low-resource NICU settings
The Williams et al. (2025) systematic review specifically noted that wired bedside monitors are expensive and often inaccessible in low-resource environments. In parts of Sub-Saharan Africa and South Asia, NICUs may lack sufficient monitoring equipment for every bed. Camera-based systems — potentially running on consumer-grade hardware — could extend monitoring coverage to settings where traditional equipment isn't available. The infrastructure requirements are a camera, a computer, and software. That's a lower bar than a conventional multi-parameter monitor that costs tens of thousands of dollars.
Where the technology still falls short
Honest assessment: several hard problems remain unsolved. Motion artifacts continue to be the biggest challenge. Neonates don't hold still. They squirm, cry, get repositioned by nurses, and undergo procedures — all of which disrupt the camera's ability to extract clean physiological signals. The Zeng et al. (2024) framework handles this better than earlier approaches, but performance still degrades meaningfully during high-activity periods.
Skin tone diversity is an unresolved concern. The rPPG signal depends on detecting subtle color changes through the skin, and melanin affects how that signal presents. Ba et al. (2023) raised equity concerns about camera-based SpO2 estimation across different skin tones, echoing well-documented problems with traditional pulse oximetry. Neonatal studies have not yet included large enough diverse cohorts to characterize this issue thoroughly.
Regulatory clearance is nowhere close. No camera-based neonatal monitoring system has received FDA clearance or CE marking for clinical vital sign measurement. The path to clearance would require large multi-center validation studies demonstrating accuracy and reliability across diverse patient populations — the kind of evidence that takes years and significant funding to generate.
What comes next
At this point, camera-based neonatal monitoring works well enough in controlled conditions that the question has shifted from "can it work?" to "can it work reliably enough for clinical deployment?" The dual-center study by Zeng et al. (2024) represents a step toward answering that, but 50 patients across two sites is still a small dataset by regulatory standards.
Near-term, expect to see camera systems deployed alongside conventional monitors in research NICUs — not replacing anything, but generating the validation data needed to push toward clinical adoption. That fits with a wider push in NICU care to reduce invasive contact with vulnerable neonates wherever possible.
Circadify has developed contactless vital sign measurement technology through rPPG and is working to bring these capabilities to clinical settings, including neonatal care. The company's camera-based platform is designed to measure heart rate, respiratory rate, and other physiological parameters without any skin contact — which aligns directly with the needs the neonatal research community has identified.
Frequently Asked Questions
Can a camera accurately measure a newborn's heart rate?
Published research shows camera-based systems can estimate neonatal heart rate with mean absolute errors between 2 and 5 beats per minute under controlled NICU conditions. Zeng et al. (2024) demonstrated robust heart rate extraction from video of 50 preterm infants, though accuracy varies with infant movement, lighting, and skin tone.
Why is contactless monitoring important for premature babies?
Preterm neonates have extremely fragile skin. Adhesive electrodes used in standard monitoring cause skin injuries in an estimated 9 to 42 percent of NICU patients. Non-contact systems eliminate direct skin contact entirely, reducing injury risk while still providing continuous vital sign data.
Is contactless NICU monitoring ready for clinical use?
Not yet as a standalone replacement for bedside monitors. Current systems perform well for heart rate and respiratory rate in controlled research settings but still struggle with motion artifacts and have limited validation for SpO2 in neonates. Most researchers position the technology as a supplemental monitoring layer.
What vital signs can camera-based systems measure in neonates?
Current research has demonstrated measurement of heart rate, respiratory rate, heart rate variability, and respiratory rate variability from neonatal video. Some systems also estimate SpO2 using multi-wavelength analysis, though neonatal SpO2 estimation is less mature than heart rate detection.