With a $1.2 million grant from Fondation Botnar, an international team of researchers will assess the feasibility of creating and launching a global-scale artificial-intelligence (AI) app for mobile devices that diagnoses diet-related problems and offers nutritional advice to adolescent girls living in urban settings in Ghana and Vietnam.
"Our hope is that a readily available AI app that is tailored to adolescents' circumstances can help them to improve their food-consumption behaviors and possibly even spillover to improving the behaviors of their peers, as well," said David Hughes, associate professor of entomology and biology, Penn State.
Aulo Gelli, principal investigator and senior research fellow at the International Food Policy Research Institute (IFPRI), said, "This project involves a new interdisciplinary collaboration between IFPRI, Penn State, the Noguchi Memorial Institute for Medical Research at the University of Ghana and the National Institute of Nutrition and the Thai Nguyen University of Pharmacy and Medicine in Vietnam, working together to extend new technology with potential to change 'business as usual' and improve the lives of millions of adolescents."
The team chose to test their app in Ghana and Vietnam because these are both lower-middle-income countries in which increasing incomes and urbanization are contributing to the adoption of poorer-quality diets, such as reliance on convenience and street foods. In Ghana, for example, these diets have led to widespread micronutrient deficiencies. An estimated 44% of adolescent girls are anemic as a result of low iron intake, among other health problems.
"Undernutrition in adolescence can have long-term consequences; it can affect their brain structure and function, and for girls, it can even affect the survival and well-being of their children," said Hughes. "Shifts to unhealthy diets and reductions in physical activity also contribute to the global increase in unhealthy weights, which, during adolescence, is associated with psychosocial problems, social stigmatization, poor self-image, non-communicable diseases in adulthood, and reduced life expectancy."
Researchers are developing an app that can recognize different foods and portion sizes to help adolescent girls living in urban settings in Ghana and Vietnam to improve their health.
In the first phase of the research, the team plans to develop new software based on existing technology created at Penn State. Called PlantVillage, the technology consists of AI that diagnoses crop diseases and nudges smallholder farmers toward improved farming practices across sub-Saharan Africa. Functionality developed specifically for the diet and nutrition project will include the ability to recognize different dishes, foods and portion sizes when the phone is held over the dish; to help users accurately record their meal intake; and to nudge users to improve their diets by providing tailored feedback based on dietary guidelines.
"The AI we're developing is truly cutting edge in two respects - it can recognize which Ghanian or Vietnamese foods are on the plate and the quantity of each food item, and it can convert raw pixels directly into nutritional content," said Pete McCloskey, lead AI engineer at PlantVillage. "We believe this will have a huge impact on the effectiveness with which nutrition experts can nudge adolescents toward healthier eating habits."
The team is undertaking focus groups with 30 adolescent girls in both Ghana and Vietnam to explore opportunities to make the new app age-appropriate, fun to use, and different and more attractive to adolescents, and to identify appropriate nudging approaches for adolescent girls to avoid unintended negative effects such as shaming. The researchers are incorporating these ideas into the app's development.
During the app development, the researchers will also assess the acceptability and feasibility of using the mobile app to track food intake and provide tailored feedback to reward and encourage healthy eating choices. The team is also preparing food atlases for the two project sites that will include photographs of different foods and mixed dishes using calibrated portion sizes. These photos will be annotated with detailed descriptions in order to provide a training set for the artificial intelligence.
The performance of the mobile app will then be validated against standard practices for dietary assessment, including weighed measurement and multi-pass, 24-hour recall, in both country settings. In addition, the team will pilot an experiment to test the effectiveness of the nudging functionality to improve healthy eating choices of the adolescent girl user group involved in the app development. The team is also planning and seeking funding for large-scale effectiveness trials of the new technology in a real-life context.
When the project is completed, the app, including the full set of functions, will be freely available to residents of Ghana and Vietnam. A version of the app with limited functionality will be freely available to users in other countries.
"If successful, this project could provide governments and other organizations with real-time diagnostics to strengthen surveillance systems and improve public programs in health and nutrition, including school meal lunch programs," said Hughes. "Ultimately, it could contribute significantly to overall public health, not only in Ghana and Vietnam, but throughout the world."
The project involves a multidisciplinary collaboration with strong in-country leadership, including Gloria Folson, Noguchi Memorial Institute for Medical Research, University of Ghana; and Phuong Nguyen and Hoang Thu Nga, National Institute of Nutrition, Thai Nguyen University of Pharmacy and Medicine in Vietnam.
Fondation Botnar is a Swiss-based foundation that champions the use of AI and digital technology to improve the health and well-being of children and young people in growing urban environments.