[ad_1]
We have had weather forecasts for decades. It is much more difficult to predict our health in the short term. Yet knowing early on that we are at risk of contracting the flu or COVID-19 could be extremely helpful. The good news: Wearable technologies, such as smartwatches, are starting to provide such early warnings.
Jessilyn Dunn is a Biomedical Engineer at Duke University in Durham, North Carolina. She was part of a team that analyzed heart rates and other data from portable devices. Smartwatch-type systems contain sensors. These collect data – a lot, a lot of it – that can indicate health or disease.
Dunn’s team asked 49 volunteers to wear wristbands loaded with sensors before and after receiving a cold or flu virus. At least once per second, these bracelets recorded heart rates, body movements, skin temperature and more. In nine out of ten recruits, these data showed signs of disease development at least a day before symptoms appeared.
Researchers describes their findings September 29 to JAMA network open.
This early warning, Dunn says, can help nip infections in the bud. This can prevent serious symptoms that would otherwise send vulnerable people to hospitals. And knowing that you are sick before you have symptoms can warn you to keep a low profile to reduce the chances of your disease spreading.
However, these systems are not yet ready for the real world, notes virologist Stacey Schultz-Cherry. She works at the St. Jude Children’s Research Hospital in Memphis, Tennessee. “It’s exciting but also very preliminary,†says Schultz-Cherry. “Much more work is needed before this approach can be scaled up. “
Sift through mountains of data
The researchers gave 31 of 49 recruits nasal drops with an influenza virus. The rest of the people were exposed to the cold virus.
Trials where volunteers agree to receive a virus are unusual, notes Schultz-Cherry. They can also be dangerous. The researchers therefore made sure that the volunteers were in good health and would not give the flu to others. (Doctors also examined them frequently during the trial.)
Dunn’s group wanted to compare sensor data from infected and uninfected people. But deciding who was infected “involved a fundamental debate within our team,” notes Emilia Grzesiak. She’s a data scientist who worked on the project while she was at Duke. The team’s final decision? Recruits were infected if they reported at least five symptoms within five days of receiving the virus. A PCR test also had to detect the virus on at least two of those days.
Recruits began to wear the bracelets before being displayed. This provided baseline data while the volunteers were in good health. The sensors continued to collect data for several days after the exposure. Some data has been measured more than 30 times per second. That means the 49 recruits each had up to 19 million data points, Grzesiak notes. A computer sifted through these mountains of data for patterns signaling an emerging disease.
For this screening, the computer needed an algorithm. Grzesiak has developed these step-by-step instructions. His algorithm tested all possible combinations of sensor data and time points. He looked for the biggest difference between infected and uninfected people. An example of a winning combo: add the average heart rate 6 to 7 hours after exposure to the virus and the average time between heartbeats 7 and 9 hours after exposure. (The best real model was more complex.)
Grzesiak used some of the data to create a computer model. She tested her predictions in the rest of the data. Then she repeated this process several times. His final model accurately predicted infections nine out of ten times.
Challenges ahead
One challenge is that many viral infections have similar symptoms. In fact, many things other than viruses trigger the same symptoms. Examples, notes Schultz-Cherry, include food poisoning, asthma, or seasonal allergies. Likewise, heart rates react to things that have nothing to do with infections. Examples include exercise and scary movies.
Also, in real life, we don’t know who has been exposed to a virus and when. Thus, this revealing post-exposure window of time will not be known. Potentially infected people can be those whose data exceeds a certain value in all two hour window. But Dunn’s team has yet to test how well the prediction model would work in this setting.
Could such a system one day indicate that people are contracting COVID-19? Maybe, says Benjamin Smarr. He is a bioengineer at the University of California at San Diego. Similar technologies, he notes, are being developed elsewhere to provide early warnings of this infection.
Such studies seem fascinating. But there is still a lot of work to be done. For example, notes Smarr, a prediction accuracy of 95% looks good. But that number means “telling one in 20 people every night that they will get the flu when in reality they won’t.”
Smarr expects continued improvements in the accuracy of predictions. Future models will likely include other types of bodily changes that indicate the development of disease. And researchers will refine these models by analyzing how well they predict effects on thousands of people.
This story is part of a series featuring news on technology and innovation, made possible through the generous support of the Lemelson Foundation.
[ad_2]
Comments are closed.