Khalifa University reveals impact of sensor pressure on health devices


A systematic scientific review led by researchers at Khalifa University in Abu Dhabi and published in npj Biosensing has shown that the contact pressure of wearable optical sensors against the skin fundamentally affects cardiovascular signal quality, creating a hidden confounder that could mislead machine learning blood pressure estimation algorithms.
The scientific review led by Khalifa University in Abu Dhabi warns that digital health algorithms may confuse true patient physiological changes with distortions caused by the tightness of a watch or sensor against the skin.
A research team led by Khalifa University in the UAE capital of Abu Dhabi published a peer-reviewed systematic scientific review on November 15, 2026, in npj Biosensing, part of Springer Nature, demonstrating that the pressure with which a wearable optical sensor contacts the skin surface can fundamentally affect cardiovascular signal quality.
The review examined 29 scientific studies addressing photoplethysmography technology and its associated blood pressure estimation methods. Although the paper does not propose a new artificial intelligence model, its core significance lies in evaluating the reliability of machine learning systems trained to interpret physiological data generated by wearable devices.
Photoplethysmography measures changes in light absorption associated with blood volume fluctuations near the skin surface. Most consumer watches and fitness bands rely on it to track heart rate, and researchers are testing its potential to estimate blood pressure without a traditional inflatable cuff via algorithms that translate pulse waveform features into physiological estimates.
The researchers found that weak contact pressure reduces optical stability and introduces motion-induced artifacts, whereas excessive pressure causes superficial blood vessel compression, alters waveform morphology, and obscures analytical features. This may lead a machine learning model to misinterpret these changes as physiological health differences rather than measurement errors caused by a hidden confounding variable.
A comparison across the 29 studies showed that favorable contact conditions varied depending on the sensor’s body location, engineering design, and targeted measurement type. Furthermore, high waveform quality did not automatically guarantee more accurate blood pressure predictions.
The available evidence did not support establishing a single universal pressure setting suitable for all cases or creating a standardized numerical ranking for machine learning methods across all devices, as comparing incompatible metrics without accounting for these physical differences yields misleading conclusions.
The review was conducted by scientists from Khalifa University in collaboration with researchers from the UK’s Newcastle University and the Massachusetts Institute of Technology in the United States, with support including a research grant from Khalifa University. It featured transparent disclosures of the authors’ research interests in biosensing and one participant’s membership in scientific editorial boards, without participating in the paper’s review decision.
The study proposes measuring, controlling, and reporting sensor pressure in a standardized, regular manner during device and model development, emphasizing that it establishes stronger foundations for scientific reproducibility and the distinction between genuine biosignals and sensor artifacts—without claiming a complete clinical product or regulatory approval for cuffless blood pressure devices.
This scientific contribution led by Khalifa University in Abu Dhabi carries striking regional and international significance in the field of health AI evaluation. Rather than promoting hasty claims that algorithms have solved the dilemma of cuffless blood pressure measurement, the systematic review reveals a fundamental physical and experimental flaw that precedes the model-building phase itself. It underscores that enhancing wearable device reliability begins with calibrating mechanical contact pressure and transparently documenting it in clinical trials so that sensitive medical predictions are not built on signals distorted by motion or superficial vascular artifacts.
The adoption of standardized contact pressure measurement protocols between laboratories and manufacturers will help accelerate the development of a more accurate and reliable generation of wearable biosensors.
Question: How does watch or sensor pressure affect the accuracy of wearable health algorithms?
Answer: Low pressure leads to motion artifacts, while excessive tension compresses superficial vessels and alters optical waveform morphology, which can mislead artificial intelligence algorithms when estimating blood pressure.
Question: Which research institutions participated in preparing this systematic review?
Answer: The study was led by researchers from Khalifa University in Abu Dhabi in collaboration with scientists from Newcastle University and the Massachusetts Institute of Technology, examining 29 published studies.
Question: Did the study determine a unified optimal pressure level for all smartwatches?
Answer: No, the review concluded that the appropriate pressure varies depending on the measurement location on the body and sensor design, and it called for standardizing pressure measurement and disclosure in research.