Smart food: Crowdsourcing of experts in nutrition and non-experts in identifying calories of meals using smartphone as a potential tool contributing to obesity prevention and management

摘要

To address an increasing global health problem of obesity, further innovative initiatives are required. One such initiative is personalized messaging using mobile applications as a potential tool contributing to obesity prevention and management. In order to achieve this, there are challenges that need to be considered first including the accurate estimation of calories of meals and individuals' calorific intakes using a smartphones. There is also a lack of evidence indicating whether novices, peers and family members can provide accurate tailored feedback on calorie intake and nutrition. The two study objectives were i. To determine the feasibility of experts in nutrition and non-experts accurately identifying calories of meals from photographs as taken on a smartphone; and ii. To inform the development a personalized messaging system for obesity prevention and management using a mobile application. This study was an experimental design using a quantitative online survey with 24 participants, consisting of 12 experts in nutrition and/or dietetics, and 12 non-experts. The non-expert group attended a training session and both groups completed an online survey. The survey consisted of 15 meals, the participants were required to view the photographs and then answer the following question for each photograph: “From viewing the above photograph, enter the number of calories you consider is in this meal? ___Kcal OR ___KJ”. Crowdsourcing was used. The results revealed that the percentage difference between the estimated calories count in the meals against the actual number of calories was on average +55% (SD 79.9) for the non-expert group and +8% (SD 15.1) for the expert group (t-test, P<;0.001). When using crowdsourcing, aggregating opinions from experts and also non-experts improves accuracy. The mode estimate from a crowd of experts is more accurate than 79% of individual experts. The crowd of non-experts' average median difference out performed 63% - f individual non-experts. Thus the crowd of non-experts is more accurate in estimating calories from photographs taken on a smartphone than most individuals. When designing a personalized messaging system for obesity prevention and management using a mobile application, a crowd of experts in nutrition and also a crowd of non-experts should be included to estimate calories in foods from photographs taken on a smartphone. This may have potential in contributing to obesity prevention and management, which warrant further research.