Skip to main content
ARS Home » Pacific West Area » Davis, California » Western Human Nutrition Research Center » Immunity and Disease Prevention Research » Research » Research Project #438876

Research Project: AI Institute for Next Generation Food Systems

Location: Immunity and Disease Prevention Research

Project Number: 2032-51530-026-018-R
Project Type: Reimbursable Cooperative Agreement

Start Date: Oct 1, 2020
End Date: Aug 31, 2022

Objective:
The proposal would fund the AI Institute for Sustainable Food Systems. The overall mission of the institute is to leverage artificial intelligence (AI) to maximize U.S. competitiveness in the breeding, production, processing and distribution of nutritious food through risk minimization and workforce development. The objective of the ARS work is to lead the Nutrition Cluster at the Institute. The goal of the nutrition cluster will be to advance AI technologies to enable precise assessment of what people are eating, quantify that food’s molecular composition, and predict the impact on their health. We will specifically address this goal with a concrete project in which food glycan content is predicted from dietary intake data.

Approach:
In collaboration with University of California (UC) Davis, we will improve artificial intelligence (AI) technologies for nutrition research in the context of a specific project to predict dietary intake and to predict the glycan composition of those foods. This consists of two parts: AI-assisted dietary intake assessment, and prediction of food glycan intake. AI-assisted dietary intake assessment: Current validated methods are text-based methods that are captured by participant recall (e.g. over past 24 hours, over past 3 months, etc). Human participants will be asked to collect real-world food diaries consisting of images and text. AI methods will be developed to convert the data from these food diaries into core food ingredients and quantities. Our preliminary data shows that natural language processing, combined with nutrient profiles, successfully maps new foods to a food composition database. Encoder/decoder convolutional neural networks show promise for the prediction of ingredients from images. AI methods will be used to convert real-world food diaries, which consist of images (meals, ingredient lists) and text descriptions, to food items and/or food codes in USDA Food Data Central.