Over half of all participants discontinue exercise programs within one year. Value placed on personal aesthetics is a powerful predictor of compliance to weight-loss regimes. Thus, providing a realistic target for body shape may improve treatment compliance. 3D optical whole-body imaging is an accessible method of body composition assessment and shape visualization. In this study, we developed data-driven statistical models to estimate body composition from 3D scans, and we derived deformable 3D shape models representing fat and muscle changes.
720 healthy adults are being recruited as part of the Shape Up! Adults cohort study. Participants undergo 3D whole-body surface scans on a Fit3D Proscanner and a whole-body DXA scan on a Hologic Horizon/A. For this analysis, 3D scans were spatially registered by an exercise physiologist using 75 fiducials. PCA and LASSO were used to develop models to predict fat mass (FM) and fat-free mass (FFM) from body shape. Age, height, FM, and FFM were mapped to body shape via manifold regression, enabling data-driven models deformable to specific changes in FM and FFM.
Data were collected on 176 healthy adults (72 male). Mean [SD] BMI (kg/m^2) was 27.7 [5.3] for men and 27.3 [6.6] for women. FM and FFM were accurately estimated from PC modes capturing 95% of 3D variation: R2 [RMSE (kg)] of 0.83 [3.67] and 0.90 [3.71] respectively for males, and 0.94 [2.38] and 0.91 [2.52] for females. Python and Blender were used to visualize participants at their current shape and for any combination of FM and FFM changes of up to ±20 kg each, representing BMI 16 to 40.
Body composition can be estimated from 3D optical scans in adults. To our knowledge, this is the first body composition model that can anticipate shape changes from target fat and muscle mass. With more Shape Up! data, this model can extend to shapes associated with diet, blood markers, and genetics. This tool may provide powerful motivation to those seeking a healthier body shape.