COVID-19 research is becoming increasingly important in helping doctors understand how the virus spreads and what we can do to slow it or stop it altogether.
Now, results from a Utah and Maryland study are giving medical experts an idea of what environments are being impacted by COVID-19 infections the most and why.
University of Utah Professor Tolga Tasdizen said he used Google Street View images to train a computer to recognize environments that contribute to COVID-19 spread.
"In this case, what we did is we took Google Street View images and created a computer algorithm to train a computer model to recognize things like: Is there a single-family housing unit in the picture? Is there a multi-family housing unit? Is there a crosswalk in this picture? Is there a sidewalk? How much green space is there in this picture?" said Tasdizen.
Before using the computer to gather this data, researchers said they would usually collect it by sending people out to talk to community members and take notes on what neighborhoods looked like.
Tasdizen said this takes a lot of effort to scale their research up to very large numbers of communities and is very challenging in terms of time and cost.
The technology lets researchers automate the whole process and scale it up to a nationwide study faster.
They started by using 18,000 Google Street View images to train a computer to recognize important environmental factors that can contribute to the spread of the coronavirus, like single-family housing vs. multi-family housing, access to green space or parks, and the number of crosswalks or sidewalks where people who are infected could come into contact with other people.
The training data was then applied to 164,000,000 Google Street View images from across the country.
"It’s quite fast. It takes about a week on a computer and then it associates the results we get with COVID-19 rates," Tasdizen said.
The data is then handed off to doctors so they can better understand which communities COVID-19 is spreading within and how environmental factors may be affecting it.
"We are expecting areas that are more densely populated would have higher rates of COVID-19 spread because of the closer proximity of people," said Tasdizen.
Researchers admit the use of Google Street View imagery may be a setback because environments may have changed since the images were last taken.
They also say their study does not show a direct link between coronavirus infection rates and the makeup of an area.
However, they say it does reveal some interesting points.
For example, a zip code with more sidewalks had 40 percent more COVID-19 cases, in this study.
Also, zip codes with more multi-family housing structures had 21 percent more virus cases.
Meanwhile, areas with more single-lane roads had a decrease in COVID-19 cases as well as areas that had more green space.
The research was funded by the National Library of Medicine of the National Institutes of Health
