Whether it’s playing a game, reading the news, checking our social media profile, snapping a quick photo or, you know, actually calling someone, over the decade-and-a-bit that smartphones have existed, they’ve fast become a one-stop-shop for just about everything we do on a daily basis. Could they soon become a comprehensive health monitoring system as well?
Researchers at the University of Toronto believe so. And they’ve developed the software to prove it. Called Anura, it’s a groundbreaking mobile app which provides users with a host of “health indexes” gleaned from nothing more invasive than a 30-second selfie. Your next question is likely some variation of, “but just how much information is it possible to gather from a selfie?” The answer: Probably a lot more than you think.
“My risk of cardiovascular diseases is at 7 percent,” mused Kang Lee, a professor of applied psychology and human development at the Ontario Institute for Studies in Education. “That’s good, but not excellent.” He was staring at his iPhone, reading off the results of his latest product demo. The screen displayed a dashboard, showing his current health profile as a series of dials; each one running from green (the best) to red (the worst.) “My diastolic blood pressure is yellow, which is not good,” he said, and frowned. “I’ve been seeing this for the last few months. I’m a little bit worried. I’m trying to bring it down to green.”
Lee brightened slightly. His heart attack risk was only 3 percent, and his risk of suffering a stroke just one percentage point higher than that. “Overall, my health assessment is 71,” he told Digital Trends. “I’m striving for 80.” Not bad for a man the app correctly predicted as being 56-years-old.
Getting under your skin
Anura promises an impressively thorough physical examination for just half a minute of your time. Simply based on a person’s facial features, captured through the latest deep learning technology, it can assess heart rate, breathing, stress, skin age, vascular age, body mass index (yes, from your face!), Cardiovascular disease, heart attack and stroke risk, cardiac workload, vascular capacity, blood pressure, and more.
This is accomplished using something called transdermal optical imaging. It’s a contactless method of gathering data by using the optical sensors in your phone to image hemoglobin changes under the skin. This, in turn, signifies psychological and psychological state changes in the user. Where features like the iPhone’s Face ID use bounced infrared beams to ascertain the surface geometry of our faces, Anura goes deeper. Literally.
“Because our skin is translucent, light can penetrate through its surface,” Lee said. “Underneath, it encounters two types of protein. One is melanin and the other is hemoglobin. When the light bounces back to our phones, our phone’s amazing optical sensors are able to pick up this information. For this purpose, we throw away the melanin information and instead analyze only the hemoglobin. We can pick up a detailed picture of how hemoglobin is concentrated in different parts of the face.”
“Getting the lighting right is very, very important.”
He pulls up a picture showing a subdermal image of the human face. It pulsates like a stormy weather map. “The brighter the colors, the higher the concentration levels,” Lee said. “You can pick up a lot of information, depending on the computational models you build.”
Each metric is gathered using a different computational model looking at the same overall information. It’s a bit like getting a room full of students to write their own stories based on a common theme. The results are impressive. In a recent paper in the American Heart Association journal Circulation: Cardiovascular Imaging, the researchers demonstrated that they could measure blood pressure with up to 96 percent accuracy.
“We think our technology can be a game-changer because it utilizes the existing technology in smartphones, which are increasingly ubiquitous,” Lee continued. “We spend a lot of time staring at our phones. If we could turn a few seconds of that into a health check, you would know if your blood pressure or other health indexes are problematic.”
The biggest challenges
Getting this to work is, of course, a challenge. In addition to training the A.I. algorithms to accurately predict the different health measures, there’s the problem of making sure the face data they have to work with is as good as possible. That involves taking over the camera controls in your smartphone and tweaking things like the ISO speed and white balance to ensure the highest quality information. Impressively, this is customized according to local environmental conditions, meaning that the app makes adjustments depending on how good the lighting is wherever you are. “Getting the lighting right is very, very important,” Lee said.
“[The goal is to] make you realize when you need to take action.”
Another issue to overcome was designing a user interface that would make sense of information like blood pressure to everyday users who don’t necessarily know whether a blood pressure of 120 over 80 is good, bad, or indifferent. (It’s pretty average, actually.) They also wanted to do this without dumbing down the information provided.
“That’s a struggle we had,” Lee acknowledged. “For each measurement, we [opted to have] green, yellow and red dials to show the results. But if you click each button, you can see the medical-level explanations about what is being measured and what it means. It also shows you what the follow-up activities are. We’re not cheapening it, or making it like a game, but making it into something that non-experts can understand. [The goal is to] make you realize when you need to take action.”
Lee and his colleagues hope that the finished product could be a useful tool for patients and doctors alike. He imagines that it could prove useful in parts of the world where people don’t have ready access to consistent high quality healthcare providers. It could also be used by the elderly or people who otherwise have difficulty leaving home for regular checkups.
But it could equally be employed by individuals who just want a more convenient, and cheaper, way to check their vital signs are reading as they should. “This is not about replacing doctors or the FDA-approved cuff-based systems for blood pressure,” Lee said. “This is a tool to help make us aware.”
A basic version of the app, measuring heart rate and stress, is available on iOS and Android. A more detailed version will hopefully in the future.
