Wearable blood pressure monitoring technology

Monitoring, understanding, and gaining a clear picture of blood pressure levels is critical for individuals’ current and future health, and soon, people may be able to keep track of these measurements daily from their finger or wrist.

For the past decade, biometric technology company Valencell has been developing innovative sensor technologies for wearables and hearables to enable impactful health outcomes. Most recently, their researchers have cultivated a calibration-free, cuffless blood pressure (BP) technology that can be used in a range of devices including smartwatches, fitness bands, patches, and pulse oximeter finger clips to provide accurate readings. The solution only requires three key parameters – a photoplethysmography (PPG) signal, motion signal, and static biometrics (height, weight, age, gender).

Unlike other monitoring methods such as electrocardiogram (ECG) or pulse transit time, this technology only needs data from PPG and inertial sensors. This could, for example, allow users to obtain precise readings throughout the day by wearing their own earbuds. The data, along with the user’s physical characteristics, are processed by the Valencell BP estimation software in an embedded model created using machine learning techniques and data from tens-of-thousands of patient results.

“We wanted to take technology that used to only work in hospital settings and place it into a device that people already wear as part of their lifestyle. We took that core layer that was designed to provide motion-tolerant heart rate and other biometrics as you live your everyday life. Then we used that core technology, and we built a layer of machine learning around it,” explains Valencell President and Co-founder Dr. Steven LeBoeuf. “That machine learning connects the dots between someone’s actual blood pressure and their blood flow information. We use this motion-tolerant PPG technology that measures changes in blood flow and blood vessels.”

The PPG sensor provides useful information for the user because it can measure more at one spot within the body compared to other mechanisms. It uses light to sense motion in the body to see how it interacts with the tissue, collecting the necessary information to determine measurements such as heart rate, respiratory rate, or blood flow.

Designing PPG sensor technology

Building the PPG machine learning-based model required a lot of data – 50,000 data sets of 5,000 subjects – collected from different parts of the body including the ear, finger, wrist, and arm. Once collected, data are processed on the device’s microchip and then sent through a telemeter to a mobile device. The first reading is sent in about 30 seconds and then blood pressure rates are updated every second.

Although PPG has long been used to measure blood flow changes and translate these changes into pulse rate, LeBoeuf explains that one major obstacle faced while expanding the technology was motion noise. PPG sensors are extremely sensitive to motion, so while in development, the team focused heavily on guiding light to the right spots in the body to eliminate noise during processing and properly filter the blood flow signal.

“The rules of the game usually used in optics didn’t apply when bodies aren’t moving, therefore, a lot of the core technology comprises guiding light on the surface where it needs to be,” LeBoeuf says. “When that light is captured by a photodetector, there’s a bunch of noise in it. We had to determine how to remove the noise dynamically in time so that processing could be the next layer.”

Through this work, Valencell was able to generate clean PPG information which they predict could be processed even further with machine learning models to make other important predictions.

“There’s a lot you can do with that clean PPG information that you can’t really do with other sensory modalities. My best guess is that in maybe 10 to 20 years, almost all medical wearable devices will have some PPG component to it,” LeBoeuf says.

LeBoeuf goes on to explain that in some cases with machine learning models, it can take that information and turn it into estimations of other important measures, such as heart rate or glucose levels.

“Some of the research we’re doing today for next-generation technologies is using that information to predict whether you’re about to have a spike in glucose or a drop in glucose,” LeBoeuf adds. “It’s not measuring the glucose level, that’s not the purpose of it. We see an opportunity to enable predictions of a spike or drop, which if you could do with a device that you already wear around anyway and it’s not invasive, that really changes the game.”

Enabling impactful health outcomes

“We get this question a lot of what’s next or what the next great sensor technology will be,” LeBoeuf explains as he sets sights on the future. “It’s really more about applying the existing capabilities at scale and then layering on the data science and the machine learning to surface new insights.”

Wearable sensors, such as those being developed by Valencell, hold huge potential in increasing value in public health, both in terms of improving health and providing more manageable costs for those in need. It’s predicted that wearables will continue to advance and could eventually be used to help indicate early stages of diseases including cancer or Parkinson’s.

“I think a lot of what wearables would do for public health is going to be in screening, as an early indicator that things could be going wrong, to provide interventions that can improve health at low cost,” LeBoeuf adds. “And it doesn’t require any scientific breakthroughs or new sensor technology necessarily, those capabilities exist today. What’s different is the aggregation of the data at scale.”

Valencell Inc.
https://www.valencell.com

About the author: Michelle Jacobson is the assistant editor of TMD. She can be reached at mjacobson@gie.net or 216.393.0323.