HCR.20.012 – Automatic dietary intake recognition for personalized dietary advice, disease prevention and health

Route: Health care research, sickness prevention and treatment

Cluster question: 075 How can we use sport, exercise, and nutrition to promote good health and what effects will this have?

Non-communicable diseases are responsible for the majority of global mortality and disability-adjusted life years. Healthy diets are important for monitoring dietary intake and (secondary) prevention. Eating is daily business, but recording it is laborious, arduous and error-prone. Making pictures is easy and making food pictures is popular. It would be a game-changer for personalized dietary advice, stimulation of a healthy dietary pattern and scientific dietary research if we could use automatic dietary intake recognition from pictures. Some techniques essential for a dietary intake method using images are already developed, but not yet working well for all foods, in real world settings and in the Dutch situation. Is for example all essential data publicly available?
We aim to develop automatic image recognition based on machine learning (ML) to collect Real World Evidence on dietary intake in a scalable, easier, attractive and affordable manner. People will share pictures of meals they eat. The pictures will be analyzed for foods, and their amounts using M-based methods. Nutrients can be calculated using (improved) existing sources. This information will be used for scientific research on diet and health, disease management, and/or returned to the individual for insight in personal eating patterns and tailor-made dietary advice.
The proposed project will be dedicated to:
– Development and validation of a machine-learning approach to quantitatively estimate dietary intake from food pictures.
– Exploring possibilities and barriers for use of pictures for nutrient intake estimation.
– Exploring possibilities and barriers for incorporation automatic dietary intake recognition in Eetmeter/National food consumption survey.
– Testing whether receiving personalized dietary advice based on pictures changes dietary behavior/dietary intake and mental/cardio-metabolic health.
– Enriching existing cohorts with nutrient intake data and performing research on diet and health.

Keywords

Automatic, Diet, health, machine learning, Monitoring, Nutrients, Personalized, Prevention

Other organisations

Eindhoven University of Technology (TU/e), Genzai, Rijksinstituut voor Volksgezondheid en Milieu (RIVM), The Netherlands Nutrition Centre, Tilburg University (TiU), Utrecht University (UU), VerdiFy

Submitter

Organisation UMC Utrecht (UMCU)
Name Prof. dr. Y.T. (Yvonne) van der Schouw
E-mail y.t.vanderschouw@umcutrecht.nl
Website www.umcutrecht.nl