Digital Single-Image Smartphone Assessment of Total Body Fat and Abdominal Fat Using Machine Learning

Figures

• 

  Figure 1. Illustration of the steps in the method of single-image, two-dimensional digital photography to characterize body shape and assess body composition: (A) image acquisition; (B) transfer to the Cloud for direct machine learning process including profile detection, extraction and conditioning of the image; (C) silhouette with body profile detect, extraction, conditioning; (D) return of silhouette with body shape and composition estimates.

• 

  Figure 2. Images showing the sequence of digital image acquisition and software-determined landmarks for estimation of body fat. Digital photograph of lateral side of an individual in an unrestricted environment. Uploaded digital image of lateral surface of the individual with software-derived silhouette showing horizontal transverse lines designated with machine learning that identify the chest, belly, thigh, and calf sites.

• 

  Figure 3. Linear regression plots of dual x-ray (DXA)-measured and two-dimensional (2D) digital imaging-estimated fat mass (FM) values in females (a) and males (b).

• 

  Figure 4. Bland–Altman plots illustrate the differences between individual dual X-ray absorptiometry (DXA)-measured and two dimensional (2D) digital image-predicted fat mass (FM) values as a function of the mean values of females (a) and males (b). Linear regression line describes the bias and 95% confidence intervals (1.96 SD) indicate the limits of agreement. Bars indicate 95% confidence intervals for mean and LOA values.

• 

  Figure 5. Linear regression plots of dual x-ray (DXA)-measured and two-dimensional digital imaging photography (DP)-estimated abdominal (L2–5) fat mass values in females (a) and males (b).