2021/04/26 | People | Artificial Intelligence
Clinical dietitian Maria Vasiloglou on Nutrition Hub
"I believe that multidisciplinary work is fundamental to create a complete system that meets user needs," says Maria Vasiloglou, PhD student at the ARTORG AI in Health and Nutrition (AIHN) lab. In an interview with Nutrition Hub, she explains how she as a dietitian works in the lab with a team of engineers and computer scientists and in close interaction with healthcare professionals to create user-friendly technology for dietary monitoring and assessment.
The AIHN lab has created an image-based app (goFOOD) that automatically translates meal images/video into nutrients with the use of artificial intelligence. It is currently being evaluated in a usability study. As a dietitian, Maria Vasiloglou mostly works on the user-perspective and clinical aspects of image-based apps.
"I was curious to gather healthcare professionals‘ opinions, who work with nutrition-related diseases, since they are the key players in recommending nutrition apps. This is why we conducted an international survey with 1001 participants from 73 countries and 6 continents. We found out that almost 46% of them have recommended nutrition apps to their clients or patients. Around a quarter of those who have not yet recommended an app, do not know of their existence!"
Another main area of research falls into the observation that major loss of data, and thus impaired quality of them, can be attributed to the common mistakes made by users of image-based nutrition apps. Users would either not include the entire plate of their food in the input picture for the nutrition app or fail to correctly estimate portion sizes in apps that require quantities as manual input. "That is where AI comes in and can solve those problems by literally estimating the portion size of the food/drink that is about to be consumed. Machine learning is a way in which hundreds of thousands of photos are gathered with the goal that the system will start learning to recognize the foods and be able to estimate the different portion sizes."