Anthropometric evaluation of a 3D scanning mobile application

Abstract

Objective: Three-dimensional (3D) imaging systems are increasingly being used in health care settings for quantifying body size and shape. The potential exists to provide similar phenotyping capabilities outside of professional settings using smartphone applications (apps). The current study aim was to compare waist, hip, upper arm, and midthigh circumference measurements acquired by a free downloadable app (MeThreeSixty; Size Stream, Cary, North Carolina) and a conventional 20-camera 3D system (SS20; Size Stream) with those measured with a flexible tape at the same anatomic sites.

Methods: Fifty-nine adults were scanned with the app and SS20; the same software was used to generate circumference estimates from device-acquired object files that were then compared with reference tape measurements.

Results: The app and SS20 had similar coefficients of variation that were minimally larger than those by the tape (e.g., waist, 0.93%, 0.87%, and 0.06%). Correlations of the app and of SS20 with tape circumferences were all strong (p < 0.001) and similar in magnitude (R² s: 0.72-0.93 and 0.78-0.95, respectively); minimally significant (p < 0.05 to p < 0.01) bias was present between both imaging approaches and some tape measurements.

Conclusion: These proof-of-concept observations combined with ubiquitous smartphone availability create the possibility of phenotyping adult body size and shape, with important clinical and research implications, on a global scale.

Figure 1.

Example of the MeThreeSixty app and SS20 scanner 3D avatars generated on a male participant. The small artifact in the participant’s right hand on the SS20 scan is from the system’s handles held during the image acquisition process. The 3D avatar on the same subject generated by the Universal Software along with demarcation of circumference sites is shown in Figure S1.

Figure 2.

MeThreeSixty app and SS20 regression analyses (A) and Bland-Altman (B) plots for four representative circumference evaluations. The regression plots include the line of identify and the generated regression line along with the line’s equation and R2. All regression analyses are p<0.001. For Bland-Altman plots, the solid horizontal line is the mean and dashed lines are ±2SD. The regression lines and their equations are presented in each panel of the figure. †, p<0.05 and ‡, p<0.01.

Figure 2.

MeThreeSixty app and SS20 regression analyses (A) and Bland-Altman (B) plots for four representative circumference evaluations. The regression plots include the line of identify and the generated regression line along with the line’s equation and R2. All regression analyses are p<0.001. For Bland-Altman plots, the solid horizontal line is the mean and dashed lines are ±2SD. The regression lines and their equations are presented in each panel of the figure. †, p<0.05 and ‡, p<0.01.