RATIONALE: Tele-dietetic is not common in current practice. To promote this, food images as dietary intake records should be validated before application. Two-dimensional (2D) digital images are reliable and have been validated in previous studies. However, the depth is virtual with a 2D image. And ingredient types, sauce types, cooking methods, and amount of oil added have not been researched for accurate food analysis. AIM: This study was designed to compare the reliability and accuracy of these parameters estimated using 2D and three-dimensional (3D) food images. METHODS: Ten nutritionists evaluated 10 selected food items between January 2008 and June 2008. Ten 2D food images and ten 3D images of the same food items were captured for the observers' evaluation. The actual weightings or volume of the food items were measured as the gold standard for comparisons. Intraclass correlations (ICCs), percentage agreement, and one-sample t-tests were analyzed to compare the reliability and accuracy of each type of images. RESULTS: Both images showed high reliability among observers with 3D images giving less variance (2D: ICC=0.916, F=17.001, p<0.001; 3D: ICC=0.846, F=6.501, p<0.001). Both images resulted in good ingredient and cooking method identification, but 3D was better in identifying sauce type. 2D images provided better volume and oil estimation when compared with 3D images. CONCLUSIONS: The research findings confirmed the application of 2D and 3D food images as reliable and accurate dietary records for nutritionists to evaluate clients' dietary habits. This implied the feasibility of tele-dietetics that one's nutrition status could be assessed over the Internet.
RATIONALE: Tele-dietetic is not common in current practice. To promote this, food images as dietary intake records should be validated before application. Two-dimensional (2D) digital images are reliable and have been validated in previous studies. However, the depth is virtual with a 2D image. And ingredient types, sauce types, cooking methods, and amount of oil added have not been researched for accurate food analysis. AIM: This study was designed to compare the reliability and accuracy of these parameters estimated using 2D and three-dimensional (3D) food images. METHODS: Ten nutritionists evaluated 10 selected food items between January 2008 and June 2008. Ten 2D food images and ten 3D images of the same food items were captured for the observers' evaluation. The actual weightings or volume of the food items were measured as the gold standard for comparisons. Intraclass correlations (ICCs), percentage agreement, and one-sample t-tests were analyzed to compare the reliability and accuracy of each type of images. RESULTS: Both images showed high reliability among observers with 3D images giving less variance (2D: ICC=0.916, F=17.001, p<0.001; 3D: ICC=0.846, F=6.501, p<0.001). Both images resulted in good ingredient and cooking method identification, but 3D was better in identifying sauce type. 2D images provided better volume and oil estimation when compared with 3D images. CONCLUSIONS: The research findings confirmed the application of 2D and 3D food images as reliable and accurate dietary records for nutritionists to evaluate clients' dietary habits. This implied the feasibility of tele-dietetics that one's nutrition status could be assessed over the Internet.