(OR09-03-23) Development of an Automated Smartphone App Feature To Accurately Estimate Food Portion Sizes

Objectives: There are several commercial weight loss apps that track dietary intake. Many of these apps are limited in their ability to capture portion sizes of food consumed, a critical element to estimating calorie and nutrient intake. We investigated the validity of weight data obtained from an in-house smartphone app (Vibroscale) with a weighing feature for food and dishware items.  

Methods: Two testing phases were conducted: 1) to establish proof of concept that Vibroscale can accurately measure weight of food items and dishware and 2) to modify and test the new Vibroscale iteration. In the first phase, Vibroscale estimated weight using the difference of phone vibration intensity with and without an item placed on the phone (i.e., intensity difference). Each food and/or dishware item measured three times on the Google Pixel 3A and Samsung Galaxy s8+ (total N =  60 observations). Heavier items are expected to have larger intensity differences. Mixed effect models examined associations between the intensity difference and mean item weight measured by a food scale (Escali San Jamar model), accounting for repeated measures. Subgroup analyses were performed to determine whether validity differed by dishware material, food item type, or phone model. In the second phase, Vibroscale was modified to present weight data - rather than intensity difference - of each item, using the Samsung Galaxy s22 (total N =  60 observations). Similar methods and analyses as Phase 1 were performed for Phase 2, except food was weighed on a ceramic plate or in its original packaging. 

Results: For Phase 1, larger intensity differences were associated with greater weight of items (β = 1340.3 ± 294.1; P< 0.001). Subgroup analyses revealed that associations between intensity differences and actual weight of items significantly differed by dishware type (p-interaction < 0.001) but not phone model nor food item (p-interactions = 0.48 and 0.24, respectively). For Phase 2, weight obtained from the phone app was not significantly associated with weight from the food scale (β = -0.01 ± 0.01; p=0.29).

Conclusions: The Vibroscale app is a promising tool to estimate food portion size. However, future app iterations must account for dishware type and for newer phone manufacturing differences that can alter phone vibration.

Funding Sources: AWL and NA receive NIH funding. This study was partially supported by the ASN UDRF Award.