ABSTRACT
Technology-assisted dietary assessment has the potential to improve the accuracy of self-reported dietary intake. This study evaluates MealScan3D (MS3D), a mobile device-based food recording system, which uses three-dimensional images to obtain food volumes and an application to capture algorithm-driven food intake data. Participants (n = 179) were randomly assigned and trained to record three meals using either MS3D or a written food record (WFR). Generous amounts of standardized meals were provided, and participants self-selected portions for each food. The weights of provided and uneaten/leftover foods were used to determine true intake. For total energy intake (three meals combined), validity (Pearson correlation) was significantly higher for MS3D vs. the WFR (p < 0.001); when interpreted as the percentage of variance in energy intake explained, MS3D explained 84.6% of true variance, a 25.3% absolute and 42.6% relative increase over the 59.3% explained by the WFR. For 9 of 15 individual foods, the Pearson correlations between true and reported portion size estimates were significantly larger for MS3D than the WFR. Bias was smaller (intercepts were closer to the means) for 9 of 15 foods and the regression coefficients for 10 of 15 foods were significantly closer to 1.0 in the MS3D arm. MS3D is feasible for dietary assessment and may provide improvements in accuracy compared to WFRs.
