2023/10/04 | Research | Artificial Intelligence
Monitoring nutrition with a single image
In a refinement of its earlier system for automated nutrient assessment via smartphone, the AI in Health and Nutrition laboratory in collaboration with the School of Health Professions, Bern University of Applied Sciences, has developed an easy-to-use single-image-input pipeline to output nutrient composition within the goFOOD(TM) application. In a retrospective real-world study, the new system reduced user burden and showed promising results.
The goFOODTM system pipeline. The previous version of our system required two images from different angles as input, while the new method requires only a single image.
(https://doi.org/10.3390/nu15173835)
The goFOODTM automatic system performs food segmentation, recognition, volume, as well as calorie and macro-nutrient estimation from single images that are captured by a smartphone. By conducting a feasibility study with 50 participants from Switzerland who recordedtheir respective meals for one day. Afterwards, dietitians carried out a 24h recall to estimate the nutrient content, as standard practice.
The collected images were analyzed retrospectively to hence assess the energy and macronutrient content of the meals. Noticeably, by comparing results with the dietitians’ estimations, the new system has comparable estimation performance with the previous method while requiring only a single-imageinput. Authors conclude that the system has thus the potential to facilitate the monitoring of individuals’ dietary habits while reducing the costs associated with dietary assessment.
