Scientists and engineers in Europe are embarking on a quest to see if they can change the way young people at risk for becoming obese eat. Key to this will be developing unobtrusive technology that monitors how quickly or slowly a person is eating and guides them toward a healthier pace.
“It’s a behavioral issue,” explains Anastasios Delopoulos, the project leader and a professor of electrical and computer engineering at the Aristotle University of Thessaloniki, in Greece. When a person begins to eat they typically begin at a high rate and slow down until they feel full. “It’s similar to the voltage of a capacitor as more and more electrons accumulate in it,” he says.
However, obese people or people at risk for the condition have difficulty feeling full, and so they tend to eat a constant, high rate. Some people at risk for eating disorders, such as anorexia, have a similar problem. But for them, the rate is unusually low. “It’s two sides of the same coin,” says Delopoulos.
Scientists at the Karolinska Institute in Sweden have already measured these rates in patients using a device called a mandometer, which was developed by AB Mando Group. The mandometer is essentially a scale that sits beneath a patient’s plate and records how quickly it lightens as the patient eats. Scientists have found that by making patients mimic a normal eating curve, they can train them to have a more normal sense of satiety—thereby treating the obesity or the eating disorder.
The challenge for the new, three-year European Union-funded project, called Splendid, is to bring that monitoring and treatment out of the clinic and into the real world. “Now we want to move toward prevention,” says Delopoulos. “We want to target some students who are not obese and identify who [among them] are at risk of becoming obese.”
For that they’ll need to develop less-obtrusive monitoring and behavioral modification technology, and the software to run it. On the hardware side of things, the Splendid researchers are working on developing wearable tech that would be able to understand and monitor chewing. The first option is to use a well-placed microphone. The idea is that the sensor would capture chewing noises and be able to interpret the rate of chewing and some information about the texture of the food. They won’t be able to tell Coca-Cola from Pepsi, jokes Delopoulos, but they should be able to tell chewy things from crispy ones or liquids.
Indeed, ear-based systems have already shown promise: Engineers at the Fraunhofer Institute of Photonic Microsystems, in Dresden, tested eight chew-detection algorithms using an in-ear microphone and recently reported 80 percent accuracy for most of them.
The other option is to adapt a photoplethysmogram—a device that detects a change in the volume of tissue by monitoring the way light is absorbed or reflected. The idea here is seeing if there is an unobtrusive spot on the body where the act of chewing produces a readable signal. The Swiss Center for Electronics and Microtechnology (CSEM), a Zurich-based partner in Splendid, is in charge of that aspect of the research.
Where on the body these sensors will go depends on the quality of the signals they achieve, says Delopoulos. “We want to be as invisible as possible,” he says. So they are investigating sensor designs that would go in the ear, sit behind the ear, or hang from a necklace, among others.
The project will also include activity monitoring. As the 2014 Consumer Electronics Show (CES) indicated, there’s already a lot of commercially available activity trackers out there.
Delopoulos’ lab itself will be in charge of “signal understanding”—figuring out things like chew rate, meal duration, and other parameters from the signals they can extract from the new wearable sensors. Once they have those signals, they’ll develop the algorithms needed to tell whether a person is at risk for becoming obese, and if they’ve already been asked to modify their behavior, how well they are doing it.
Getting all that into an unobtrusive wearable device wouldn’t have been easy five years ago, say Delopoulos. Android smartphones are now powerful enough to run the needed statistical learning algorithms. And those algorithms themselves are “much more mature now,” he says. “That’s due to research carried out in multimedia indexing and retrieval.”