Researchers investigating diet and weight use food records in which the research subjects input the food they eat. However, underreporting and mis-reporting food intake is a problem that affects research outcomes. An ideal mobile food record would automatically identify and code food intake in a standardized manner from an image, but automated image recognition of food remains extremely challenging and rudimentary and is currently unable to be deployed in a “field-ready” strategy.

Professor Lisa Chow (Medicine), Professor Lisa Harnack (Epidemiology and Community Health), and Associate Professor Rui Zhang (Surgery; MSI PI) are working on a project called “AI driven modifications of a mobile food record to enhance capture of dietary intake,” that seeks to identify needed modifications, such as addition of AI-driven prompts to ensure food detail, needed for nutritional classification of foods captured in a mobile food record. They believe that a mobile food record can be enhanced by AI-driven suggestions for free-text entry to enhance capture of dietary habits across a broad population, especially populations who have been traditionally unrepresented in behavioral weight loss studies. They are developing a machine learning-based approach to determine the extent of matching and the confidence level of matching relative to the University of Minnesota Nutrition Coordinating Center database. They will identify key missing “nodes” that will enhance the confidence of matching and reliability of reporting.

This project recently received a UMII Seed Grant. UMII Seed Grant funds are intended to promote, catalyze, accelerate and advance UMN-based informatics research in areas related to the MnDRIVE initiative, so that UMN faculty and staff are well prepared to compete for longer term external funding opportunities. 

This Seed Grant falls under the Global Food research area of the MnDRIVE initiative.

Image description: AI-driven modifications of a mobile food record to enhance capture of dietary intake.