Those who are counting calories must rely on either clunky food scales or inaccurate guesses of portion sizes. But a new wearable computer called the eButton, which matches images against a library of 3D geometric shapes, may soon provide a stealthy, accurate way of measuring calories.
The eButton – which fastens to a shirt like an ordinary badge – consists of a low-power central processing unit (CPU), a random-access memory (RAM) communication interface and a Linux or Android operating system.
The device works like a fly on the wall, automatically taking pictures of the wearer’s food and calculating calories. The creators say unlike camera-enabled cell phones, which are able to collect digital images of food only by intruding on normal life, the eButton is able to capture the images without interfering.
The researchers, who are from the University of Pittsburgh in partnership with researchers from China, studied the effectiveness of their eButton. The results were recently published online in the journal Measurement Science and Technology.
In order to estimate food volume from a single image of food on a dining plate, the device uses three processes:
- The food location is calculated using a coordinate system and the plate as a scale reference.
- The food is “segmented” from the background in the image. Image features, such as color contrast, color similarity and curve bending degree are taken into account.
- The 3D shape of the food (for example, an ellipse for a potato or a wedge for pizza) is selected from a shape model library, and the object from the image is compared with it.
Once the shape is determined and compared, the food item’s overall 3D size is determined.
The researchers note that experimental tests on the device’s ability to identify food shapes whose volume was known showed that the device “demonstrated satisfactory performance.”
Food databases, such as the Food and Nutrient Database for Dietary Studies (FNDDS), are then used to determine the calories and nutrients in the food by sending food name and portion size data.
The researchers tested the eButton on 17 foods, including jelly, broccoli, hamburgers and peanut butter.
There were three foods that gave the researchers problems: ketchup, haddock and ice cream. They note that the geometric properties of these items resulted in the largest estimation errors. In the case of the ketchup, it was too small, whereas the ice cream and haddock had concave surfaces, which led to an overestimation of their volumes.
However, results show that for most foods, the eButton method had an average error of only 3.69%. The researchers note this error is much lower than that made by visual estimation, which resulted in an average error of 20%.
Although the researchers can be pleased with their results, they do note that there are still some “unresolved issues.” Variations of cooking habits and differing ingredient types across human populations, for example, could cause errors in food volume estimation, they say.
Additionally, the device has some problems when the color of the food matches that of its background – for example, rice on a white plate.
Still, they are optimistic about how the use of their device could become a tool for overweight and obese people to monitor their caloric intake, in addition to potential benefits for patients with diabetes, certain cancers and cardiovascular diseases.
Dr. Mingui Sun, co-author of the study and professor of Neurology Surgery at the University of Pittsburgh, told Medical News Today about wider uses for the device:
“Besides the described application, eButton can evaluate people’s lifestyle and help them adopt a healthy one. We are also doing research in using it to help the blind find their ways, monitor children’s food intake [and] study gain and balance for the elderly to prevent falls.”
He also said that there are non-medical applications for the device, including use by the police or military, or as an astronaut badge.
The US National Institutes of Health (NIH) supported the team’s research.