Purdue researchers use AI and wearable sensors to better understand eating habits
Purdue University researchers are combining artificial intelligence, wearable sensors and nutrition science to explore a more accurate and timely way to understand what people eat and when they may be most open to making healthier choices.
The Sensor-based Estimation of Nutrition and Surrounding Environment, or SENSE, study brings together researchers from Purdue’s Elmore Family School of Electrical and Computer Engineering and College of Health and Human Sciences. The project is led by Fengqing Maggie Zhu, associate professor of electrical and computer engineering, in collaboration with Heather Eicher-Miller, professor of nutrition science.
Traditional dietary assessments often ask people to record or remember what they ate. Those methods can be time-consuming, and participants may forget foods, misjudge portion sizes or report less than they consumed.
SENSE examines whether technology can collect useful information as eating happens. In the study, 118 participants used several tools over 24 hours, including a small sensor attached to eyeglasses. The device, known as the Automatic Ingestion Monitor, was developed in partnership with Edward Sazonov, Cudworth Professor of Engineerg at The University of Alabama. It uses a chewing sensor to trigger a camera that captures images from the wearer’s point of view during meals.
Participants also used Purdue’s mobile Food Record (mFRTM) app to take before-and-after meal photos, completed a standard dietary recall the following day and answered questions about whether they would be receptive to nutrition guidance at different moments.
“Our goal is to move beyond relying entirely on someone’s memory of what they ate,” Zhu said. “By combining wearable sensing, meal images and artificial intelligence, we can begin to understand eating behaviors in their real-world context. That could eventually allow digital health tools to provide useful support at the moment a person is most likely to benefit from it.”
Zhu’s Video and Image Processing Laboratory, or VIPER, develops computer vision and machine learning methods that help computers identify foods and estimate amounts from meal images. The laboratory created the Technology Assisted Dietary Assessment (TADATM) system, which includes the mobile Food Record app and related image-analysis tools.
Eicher-Miller and her team contribute expertise in dietary patterns, nutrition assessment and human behavior. The researchers are particularly interested in just-in-time adaptive interventions — personalized prompts or guidance delivered when an individual appears ready and able to respond.
“Improving nutrition is not simply a matter of knowing what someone ate,” Eicher-Miller said. “We also need to understand the circumstances surrounding an eating decision and whether that person is receptive to support. Bringing nutrition science together with sensing and artificial intelligence gives us a new way to study those factors as they occur in daily life.”
The long-term vision is technology that can support both more dependable dietary research and personalized nutrition tools. Unlike many commercial food-tracking apps, Purdue’s system is being developed through peer-reviewed research, using carefully curated data and measurable methods to estimate nutritional information.
The National Cancer Institute, part of the National Institutes of Health, is funding the project.