NUS Study Uses Wearable Pulse Signals to Predict Vascular Health from Overnight Sleep Data

Singapore's National University of Singapore (NUS) is at the forefront of innovative health research, with a groundbreaking study from its Yong Loo Lin School of Medicine demonstrating how everyday wearable devices can predict vascular health using simple overnight data. Researchers at the Centre for Sleep and Cognition have shown that photoplethysmography (PPG) signals—optical measurements of blood volume changes in the microvascular tissue—captured by a consumer ring like the Oura Ring during sleep can accurately estimate vascular age. This metric reflects the physiological age of a person's arteries, serving as a powerful indicator of cardiovascular risk far beyond chronological age.

Vascular age provides a snapshot of arterial stiffness, a key precursor to conditions like hypertension, heart attacks, and strokes. As arteries stiffen with age or due to lifestyle factors such as poor diet, smoking, or inactivity, pulse waves travel faster, altering waveform shapes. Traditional assessments require clinical visits with tools like carotid ultrasound or tonometry, limiting frequency. The NUS approach leverages passive sleep monitoring, where motion artifacts are minimal, turning nightly routines into opportunities for proactive health insights.

The Science of PPG and Pulse Waveforms

Photoplethysmography works by shining light through the skin to detect pulsatile blood flow. In wearables, green or infrared LEDs illuminate the finger, capturing reflections or transmissions as voltage changes synchronized with heartbeats. Key waveform features include crest time (from pulse onset to systolic peak), diastolic time (systolic to diastolic peak), and reflection index (reflected wave amplitude relative to systolic). These evolve with age: older vessels show prolonged crest times and heightened reflections due to reduced elasticity.

The NUS team processed overnight PPG from 160 healthy adults aged 20-70, filtering signals and extracting over 370 pulses per person. Despite subtle differences—ring signals had rounder systolic peaks—they correlated highly (median 0.97) with clinical fingertip probes. This reliability during sleep positions wearables as viable for longitudinal tracking, potentially alerting users to early stiffening before symptoms arise.

Methodology: From Sleep Lab to Machine Learning

Participants underwent supervised overnight polysomnography (PSG) in NUS labs, wearing both a clinical-grade fingertip oximeter (256 Hz) and Oura Ring Gen3 (50 Hz). Data underwent artifact removal, 30-second segmentation, and fiducial point detection for feature extraction. Two models were tested: a linear regression on handcrafted features (crest time, diastolic time, reflection index) and a deep convolutional neural network (CNN) on raw waveforms resampled to 200 points.

The CNN, with two convolutional layers and dropout for robustness, excelled, achieving mean absolute errors of 6.28 years (clinical) and 7.25 years (ring), with correlations of 0.84 and 0.80 to actual age. Visualization via Grad-CAM revealed the model's focus on onset-to-dicrotic notch regions, mirroring age-related changes. Higher estimated vascular age also linked to elevated systolic blood pressure, validating clinical relevance. For full details, explore the published study in PLOS Digital Health.

Key Findings: Accuracy and Real-World Promise

The study confirmed wearables match clinical tools: feature correlations were strong (e.g., crest time r=0.62 for ring), and deep learning mitigated device variances. Errors were comparable across ages, though slight biases existed (underestimation in older adults). Ring-derived vascular age in the highest tertile associated with 8 mmHg higher systolic BP, underscoring predictive power.

These results build on prior NUS work linking sleep PPG changes to vascular elasticity, advancing from correlations to actionable age estimation. In Singapore, where cardiovascular diseases claim over 12,000 lives yearly amid rapid aging, such tools could transform population screening.

Photo by David Trinks on Unsplash

NUS Centre for Sleep and Cognition: A Hub of Innovation

Led by Professor Michael W. L. Chee, the Centre integrates neuroscience, sleep science, and digital health. Co-first author Dr. Gizem Yilmaz highlighted: "Signals collected passively during sleep can be translated into clinically meaningful insights about vascular health." This study exemplifies NUS's translational focus, bridging lab data to consumer tech.

The centre's prior research validated wearables against actigraphy for sleep metrics and explored travel's vascular impacts. Collaborations like the January 2026 Oura-NUS Joint Lab accelerate personalized preventive health, aligning with Singapore's Smart Nation vision. Read more on the NUS announcement.

Overcoming Traditional Limitations

• Invasiveness: No needles or cuffs needed; passive collection.
• Accessibility: Devices cost under S$500, vs thousands for clinical setups.
• Frequency: Nightly vs annual checkups.
• Scalability: Enables big data for epidemiology.

Challenges remain: contact pressure variability affects reflection index; models need diverse training (e.g., elderly, diseased). Yet, sleep minimizes artifacts, outperforming daytime use.

Implications for Singapore's Healthcare Landscape

Singapore faces rising cardiovascular burdens: 1 in 3 adults has hypertension, per MOH data. Wearables could integrate into Healthier SG incentives, prompting lifestyle tweaks like exercise or diet via app nudges. For universities, this spurs biomedical engineering programs; NUS offers related courses in bioengineering and data science.

Population studies could track interventions' vascular impacts, reducing S$1.5 billion annual CVD costs. Early detection via vascular age might avert 20-30% of events through timely management.

Collaborations Driving Progress

NUS-Oura partnership exemplifies industry-academia synergy. Oura's Cardiovascular Age feature, validated here, empowers users. Garmin and A*STAR ties expand wearables to fatigue biomarkers, positioning Singapore as Asia's health tech leader.

Photo by Artem Beliaikin on Unsplash

Future Directions and Challenges

Next: diverse cohorts, clinical validation against ultrasound-derived vascular age, integration with AI for risk scores. Ethical considerations include data privacy under PDPA. NUS aims for real-world trials in high-risk groups.

Actionable insights: Track sleep with validated rings; aim for vascular age matching chronological; consult doctors on elevations. Lifestyle: 150 min weekly exercise, Mediterranean diet, 7-9 hours sleep.

Singapore Universities Leading Wearable Health Tech

NUS joins NTU's wearable fatigue sensors and SMU's AI health analytics. Government grants via NRF fuel this, creating jobs in medtech. For aspiring researchers, NUS PhD programs in biomedical data science offer entry.