All posts by Brian T. Horowitz

COVID’s Unlikely Offspring: The Rise of Smartwatch as Illness Detector

Post Syndicated from Brian T. Horowitz original https://spectrum.ieee.org/tech-talk/consumer-electronics/gadgets/covid-byproduct-smartwatch-increasingly-illness-detector

Smartwatches and wearable devices have come a long way beyond counting steps. They’re getting better at detecting illness, and now they can even spot signs of COVID-19 before you are aware that you’re sick. Could a few data points gathered from consumer tech on your wrist really give doctors the data they need to diagnose you with a serious illness like COVID-19? 

Research is still in its early stage, but the last several months have seen a number of research efforts to increase the smartwatch’s illness detection capabilities. And it now looks like these tools will likely outlast the present pandemic.

Scripps Research has introduced an app called MyDataHelps as part of a study that tracks changes to a person’s sleep, activity level or resting heart rate. Fitbit is also building an algorithm that can detect COVID-19 before a person experiences symptoms. Meanwhile, Stanford Medicine researchers have developed a smartwatch alert system that can work on any wearable device, including Fitbit, Apple Watch and Garmin watches.

Michael Snyder, professor and chair of the Department of Genetics and director of Stanford Center for Genomics and Personalized Medicine at Stanford University, says watches can pick up signals of respiratory illnesses, even with asymptomatic cases. As COVID-19 hit, Snyder’s research increased “full blast,” he said. 

The alert system Snyder’s team ultimately developed is a “distributed system up on Google Cloud that reads people in real time,” he said. “And when it’s following their heart rate, if we see it jump up, we ping them back.” At that point researchers tell participants to stay home because they may be infected. The app uses an algorithm that warns patients using yellow or red alerts. Snyder says these alerts can help reduce COVID-19 transmission rates. In fact, 70% of the time a signal indicates a COVID-19 infection before or at the time it’s detected, he said. Snyder explained that just like a car dashboard monitors car health, a smartwatch will track a person’s physiology. 

“People like us will build algorithms that that will do the interpretation like pathologists when they see an imaging section for cancer diagnosis. They count on the pathologist to tell them what’s going on,” Snyder said. “That’s the way this will work.” 

Another study from Snyder’s group, published in a 2018 issue of the journal Personalized Medicine, examined how wearable devices like smartwatches could prevent evidence of inflammation or predict cardiometabolic health. 

For people with respiratory conditions, their heart rate jumps when they are breathing harder, Snyder noted. Another indicator that triggers alarms in Snyder’s algorithm is skin temperature. He has also developed a signal to detect signs of diabetes from drier skin. Snyder’s team is also working on expanding the types of data types it can study—and the frequency with which they sample the data. Higher frequency or resolution sampling increases the sensitivity of a data set. 

However, Snyder added, “You don’t want to oversample, because you’ll just drain your battery on the watch.”

Mount Sinai Studies How Wearables Indirectly Detect Inflammation

In another research project, Mount Sinai Health System is using the Apple Watch to detect inflammation based on changes in heart rate. Rob Hirten, assistant professor of medicine at the Icahn School of Medicine at Mount Sinai, is a gastroenterologist and had been using wearable devices to predict flares of inflammation from Crohn’s disease and ulcerative colitis. 

“When you have inflammation develop in the body in particular, you can see fluctuations in your nervous system function,” Hirten explained. When the COVID-19 pandemic hit, Hirten and his team applied the research to detect increased risk of COVID infection in healthcare workers. The custom application Hirten designed allowed his team to collect physiologic data for analysis and ask short questions about symptoms.

“What we’re able to then do through our app is take the data from the phone and collect it into our research portal so that we’re actually able to analyze the data that’s being collected by the Apple Watch normally,” Hirten said. 

Smartwatches and Biomarkers

At Duke University, Jessilyn Dunn, Ph.D. assistant professor of biomedical engineering, biostatistics and bioinformatics at Duke University, and her team are testing the role of smartwatches as another tool to use to detect physical signs of COVID-19. Dunn formerly collaborated with Snyder at Stanford on research prior to the pandemic. With one of her doctoral students studying in Wuhan, China, and providing firsthand info on the start of the health crisis, Dunn knew COVID would be an opportunity to do some research on illness detection with smartwatches. 

“We had daily electronic surveys asking people about their symptoms, testing status, those sorts of general questions to get a feel if people are sick or not, and then pulling in their smartwatch data,” Dunn said. “We could pull in data from pretty much any consumer wearable device that is compatible with an Apple phone or an Android phone,” Dunn said. 

Dunn’s team has partnered with organizations in North Carolina to distribute Apple Watches and Garmin watches to communities of color to balance the demographics of the study. 

“We’ve seen that COVID is disproportionately affecting communities of color, and so to only develop what we call digital biomarkers on the people who already own smartwatches just seemed like it was missing the mark,” Dunn said. “Equitable digital biomarker development” has been a focus of her study to overcome bias that exists in machine learning, she said.  

“We’re going to need to collect data in people who have certain types of illnesses and to be able to build these digital biomarkers,” Dunn said. The biomarkers could distinguish between an illness like diabetes and respiratory illness. In addition to heart rate, Dunn is using algorithms that can monitor blood oxygen (SPO2) levels.

The Future of Wearable Illness Detection

So can a few noisy alarms from someone’s wrist really tell you if you are sick? 

“In my lab, we’re always a little bit skeptical of making sure that technologies are properly validated,” Dunn said. She noted that consumer devices such as smartwatches were made for primarily entertainment or hobbies and less for medical purposes. However, Dunn is optimistic that the smartwatches will be able to help detect when people are sick, but oversight of the data will be key. 

“It’s really important that there is appropriate oversight set up for this sort of digital health arena,” Dunn said. “[We need to make] sure that when we’re getting information from these devices that it’s trustworthy.” 

Hirten said his team is planning to integrate additional physiological markers into their algorithms to predict additional diseases. Combining heart rate along with blood-oxygen levels and steps increases the predictive ability for medical professionals, he noted. 

“What we’re learning from the COVID infections by using these wearable devices I think will translate in the future over the coming years as we take what we’ve learned here and start applying it to other conditions and diseases to try to impact them,” Hirten said. 

Snyder says his group’s heart rate alert studies have established they can detect signs of illness.

“It’s really clear your heart rate jumps up when you’re ill,” Snyder said. The next step will be to expand to data types for other illnesses beyond the COVID-19 pandemic.