SHiFT Initiative: Using innovative AI mobile technology for diet assessment and provide tailored “nudging” on dietary intake as a strategy to improve diets in Vietnam

Background

Suboptimal diets are major contributors to micronutrient deficiencies, undernutrition, and associated morbidity and mortality, especially affecting children’s physical and psycho-social development. Nutritional attention throughout childhood and adolescence is crucial to ensure that children can thrive over the 8,000 days spanning from infancy to adulthood, and to protect investments made earlier in the life cycle.

School feeding programs, reaching over 400 million children globally with an annual investment of over $45 billion, serve as a multifaceted intervention benefiting education, health, and nutrition. Despite these investments, there is limited data on the diets and nutrition status in school age children, particularly in low- and middle-income countries (LMICs). Additionally, monitoring the quality of school meal programs in LMICs remains challenging.

This study builds on the Nudging for Good (NFG) Project, an ongoing successful collaboration that began in 2020 between Thai Nguyen University Medicine and Pharmacy (TUMP), Thai Nguyen National Hospital (Vietnam), Thai Nguyen Provincial A Hospital (Vietnam), the National Institute of Nutrition (Vietnam) and the International Food Policy Research Institute (IFPRI) (USA). The NFG project developed the PlantVillage Food Recognition Assistance and Nudging Insights (FRANI) app that offers a cost-effective solution for recognizing foods, providing consumption statistics, and estimating nutrient intakes in school children, adolescents and youth with accuracy comparable to professional dieticians but at a fraction of the cost. In the original validation study in Vietnam, we found that both FRANI and 24HR accurately estimate nutrient intake in adolescent girls aged 12-19y. Two further validation studies of FRANI conducted in Ghana, including school age children aged 9-15y and youth aged 18-24y, confirmed that FRANI-assisted dietary assessment accurately estimates nutrient intake and performed as accurately as 24HR in these age groups. Based on these results, the NFG project developed a version of FRANI that can be used to provide real-time monitoring data on the quality of school meals.

Specific tasks

Improvement of FRANI food records and calibration of FRANI portion estimation

This step involves increasing the size of the image database for the AI recognition model as well as fine tuning the AI model to improve food recognition. This includes the food recognition models for Ghana, Malawi, Sri Lanka and Vietnam.

The specific activities include:

• Updating food databases to cater for foods used in school meals in Vietnam.
• Fine tuning AI food models in Vietnam.
• Support the revision of the Standard Operating Procedures (SOPs) for calibration of FRANI portion estimation
• Undertake FRANI calibration using repeated model inference using different weight coefficients
• Provide performance benchmarks for food recognition and portion size estimation based on calibration data set
• Update FRANI food records based on calibrated model

Technical support and training on FRANI-related activities

This step involves supporting the team in Vietnam to undertake the new project activities, as well as present the project at the Global Child Nutrition Forum in Tokyo in December 2024.

The specific activities include:

• Remote technical support in updating SOPs and related FRANI procedures and databases for calibration and validation studies
• In-person training visit and presentation at GCNF in Tokyo

Deliverables:

Calibration SOPs, data set and technical support visit and trip to Japan

Duration: October 15-December 31, 2024

Required qualifications of the survey firm:

• Demonstrated experience in developing semantic segmentation models for FRANI food recognition
• Demonstrated experience in calibrating FRANI assisted food recognition and portion estimation
• Experience in using FRANI assisted diet assessment in school age children
• Experience in working with FRANI for diet assessment and school meal monitoring

Application documents:

Please include the following (in English) in your application:

• Cover letter
• Technical proposal with a detailed budget