More than 70 years ago, researchers took the first photo of Earth from space. Now, that grainy, monochrome picture has been replaced by thousands of satellite images of the world’s neighborhoods, all the way down to dogs lounging in driveways.
It turns out that satellite photos are far more useful than we could have imagined. In a recent study from the University of Washington, scientists deployed artificial intelligence algorithms to scan 150,000 Google Maps satellite photos from six cities in search of obesity.
“No one ever thought we should shoot up satellites to say something about obesity. That’s what’s so cool about this,” said Marcel Salathé, head of the Digital Epidemiology Lab at Switzerland’s École Polytechnique Fédérale de Lausanne. “It captures the imagination. New technology is doing things that we thought was impossible before.”
In the old days — as in, just a few years ago — researchers studying areas like the link between overweight citizens and their home environments would likely have to drive around and mark down what they saw. But this time around, they decided to use a deep learning technique called a convolutional neural network to answer their questions.
The algorithm identified neighborhood characteristics in the satellite images — things like the prevalence of green parks, sidewalks, and highways. Then scientists used statistical analysis to find links between these neighborhood elements and the obesity rate in cities like Seattle, Dallas, Memphis, and Los Angeles.
“This is the first time this has been done for public health,” said Elaine Nsoesie, one the study’s authors and a researcher at Boston University’s School of Public Health. “Our study shows that the features are very strongly related and correlated to obesity rates in those neighborhood. . . For example, if you live in a neighborhood where you don’t have parks and sidewalks, you’re less likely to go out walking.”
Nsoesie’s findings could influence how neighborhoods are designed in the future, but the true golden nugget is in the study’s methods. Scientists have already used similar methods to estimate poverty in African countries, and the future research possibilities could be enormous.
“From these images — from the way the humans shaped their environments — we can say things about poverty, public health, about wealth,” Salathé said. “This is the kind of surveillance that people can actually like. There’s satellites up there not only to spy on us, but to help us lead better lives.”
Kelly Kasulis is a freelance journalist. Follow her on Twitter: @KasulisK.