Madelyn K. Simmang, Katherine Sullivan, Casey J. Metoyer, Jacob Broeckel, Andrew D. Fields, Mary Lovelady, Maddy Schwing, Michael V. Fedewa, Michael R. Esco, FACSM. University of Alabama, Tuscaloosa, AL.

BACKGROUND: A smartphone application has been previously validated to estimate metrics of body composition (%Fat) from a full-body digital image. However, the reliability of the automated image analysis program has not been extensively examined under varying lighting conditions. PURPOSE: The aim of this study was to evaluate the reliability of %Fat estimates measured under Low(LL) Ambient(AL), Moderate(ML), and Bright-Light(BL) conditions. METHODS: A convenience sample of participants were included in the study (n=12, 83.3% female, 83.3% Caucasian, 31.25±10.49 yrs, 24.82±2.85 kg/m2). Age, gender, and race were assessed via self-report. Full-body digital images were taken in front of a white photography backdrop under LL, AL, ML, and BL lighting conditions (<50 Lux, 300-400 Lux, 600-800 Lux, and >900 Lux, respectively). Images were taken from the posterior view and were captured using an iPad Air 2 (Apple Inc., Cupertino, CA). A light meter (MT-912, Shenzhen Flus Technology Co., Ltd., Shenzhen China) was used to measure the level of illuminance in Lux. Images were analyzed using an automated smartphone application (made Health and Fitness LLC, Birmingham AL. version 1.1.3), which provided estimates of %Fat using a proprietary algorithm. A repeated measures ANOVA was used to assess potential mean differences in %Fat across the four lighting conditions, with the reliability assessed using a 2-way ICC with absolute agreement. The strength of the ICC value was considered weak, moderate, strong, or near-perfect (r=0.2, 0.5, 0.8, and 0.9 respectively). Data are presented as mean±standard deviation, with statistical significance set at p<.05. RESULTS: Significant differences were observed across conditions (p=.047), such that %FatLL (27.62±4.95 %Fat) was slightly higher than the %FatBL (26.94±5.44 %Fat) (p=.018), but not different than the %FatAL (27.16±5.08 %Fat) or %FatML (27.31±5.23 %Fat) conditions (both p>.05). No other differences were observed between conditions (all p>.05). Near-perfect agreement between %FatLL and the %FatAL, %FatML, and %FatBL conditions (ICC=0.984, 0.985 0.991, respectively; all p<.001) was observed. CONCLUSION: Based on the results of the study, a small difference was observed between %Fat estimates obtained under LL and BL conditions. However, the agreement between all conditions was near-perfect. These results suggest that %Fat can be estimated from a single digital image using a smartphone application across various lighting conditions with acceptable reliability.

This document is currently not available here.