A device-agnostic shape model for automated body composition estimates from 3D optical scans

Strengths and Limitations of BMI in the Diagnosis of Obesity: What is the Path Forward?

PURPOSE OF REVIEW

This review aims to discuss strengths and limitations of body mass index (BMI) in diagnosing obesity, the use of alternative anthropometric measurements, and potential new technology that may change the future of obesity diagnosis and management.

RECENT FINDINGS

The diagnosis of obesity requires the anthropometric assessment of adiposity. In clinical settings, this should include BMI with confirmation that elevated BMI represents excess adiposity and a measure of fat distribution (i.e., waist circumference (WC), waist to height ratio (WHtR), or WC divided by height0.5 (WHR.5R). Digital anthropometry and bioelectric impedance (BIA) can estimate fat distribution and be feasibly employed in the clinic. In addition, the diagnosis should include a clinical component assessing the presence and severity of weight-related complications. As anthropometric measures used in the diagnosis of obesity, BMI is generally sufficient if confirmed to represent excess adiposity, and there are advantages to the use of WHtR over WC to assess fat distribution. BIA and digital anthropometry have the potential to provide accurate measures of fat mass and distribution in clinical settings. There should also be a clinical evaluation for the presence and severity of obesity complications that can be used to stage the disease.

Mobile phone applications for 3-dimensional scanning and digital anthropometry: a precision comparison with traditional scanners

BACKGROUND

The precision of digital anthropometry through 3-dimensional (3D) scanning has been established for relatively large, expensive, non-portable systems. The comparative performance of modern mobile applications is unclear.

SUBJECTS/METHODS

Forty-six adults (age: 23.3 ± 5.3 y; BMI: 24.4 ± 4.1 kg/m2) were assessed in duplicate using: (1) a mobile phone application capturing two individual 2D images, (2) a mobile phone application capturing serial images collected during a subject's complete rotation, (3) a traditional scanner with a time of flight infrared sensor collecting visual data from a subject being rotated on a mechanical turntable, and (4) a commercial measuring booth with structured light technology using 20 infrared depth sensors positioned in the booth. The absolute and relative technical error of measurement (TEM) and intraclass correlation coefficient (ICC) for each method were established.

RESULTS

Averaged across circumferences, the absolute TEM, relative TEM, and ICC were (1) 0.9 cm, 1.5%, and 0.975; (2) 0.5 cm, 0.9%, and 0.986; (3) 0.8 cm, 1.5%, and 0.974; and (4) 0.6 cm, 1.1%, and 0.985. For total body volume, these values were (1) 2.2 L, 3.0%, and 0.978; (2) 0.8 L, 1.1%, and 0.997; (3) 0.7 L, 0.9%, and 0.998; and (4) 0.8 L, 1.1%, and 0.996, with segmental volumes demonstrating higher relative errors.

CONCLUSION

A 3D scanning mobile phone application involving full rotation of subjects in front of a smartphone camera exhibited similar reliability to larger, less portable, more expensive 3D scanners. In contrast, larger errors were observed for a mobile scanning application utilizing two 2D images, although the technical errors were acceptable for some applications.

3D Convolutional Deep Learning for Nonlinear Estimation of Body Composition from Whole-Body Morphology

Total and regional body composition are strongly correlated with metabolic syndrome and have been estimated non-invasively from 3D optical scans using linear parameterizations of body shape and linear regression models. Prior works produced accurate and precise predictions on many, but not all, body composition targets relative to the reference dual X-Ray absorptiometry (DXA) measurement. Here, we report the effects of replacing linear models with nonlinear parameterization and regression models on the precision and accuracy of body composition estimation in a novel application of deep 3D convolutional graph networks to human body composition modeling. We assembled an ensemble dataset of 4286 topologically standardized 3D optical scans from four different human body shape databases, DFAUST, CAESAR, Shape Up! Adults, and Shape Up! Kids and trained a parameterized shape model using a graph convolutional 3D autoencoder (3DAE) in lieu of linear PCA. We trained a nonlinear Gaussian process regression (GPR) on the 3DAE parameter space to predict body composition via correlations to paired DXA reference measurements from the Shape Up! scan subset. We tested our model on a set of 424 randomly withheld test meshes and compared the effects of nonlinear computation against prior linear models. Nonlinear GPR produced up to 20% reduction in prediction error and up to 30% increase in precision over linear regression for both sexes in 10 tested body composition variables. Deep shape features produced 6-8% reduction in prediction error over linear PCA features for males only and a 4-14% reduction in precision error for both sexes. Our best performing nonlinear model predicting body composition from deep features outperformed prior work using linear methods on all tested body composition prediction metrics in both precision and accuracy. All coefficients of determination (R2) for all predicted variables were above 0.86. We show that GPR is a more precise and accurate method for modeling body composition mappings from body shape features than linear regression. Deep 3D features learned by a graph convolutional autoencoder only improved male body composition accuracy but improved precision in both sexes. Our work achieved lower estimation RMSEs than all previous work on 10 metrics of body composition.

Generative deep learning furthers the understanding of local distributions of fat and muscle on body shape and health using 3D surface scans

BACKGROUND

Body shape, an intuitive health indicator, is deterministically driven by body composition. We developed and validated a deep learning model that generates accurate dual-energy X-ray absorptiometry (DXA) scans from three-dimensional optical body scans (3DO), enabling compositional analysis of the whole body and specified subregions. Previous works on generative medical imaging models lack quantitative validation and only report quality metrics.

METHODS

Our model was self-supervised pretrained on two large clinical DXA datasets and fine-tuned using the Shape Up! Adults study dataset. Model-predicted scans from a holdout test set were evaluated using clinical commercial DXA software for compositional accuracy.

RESULTS

Predicted DXA scans achieve R2 of 0.73, 0.89, and 0.99 and RMSEs of 5.32, 6.56, and 4.15 kg for total fat mass (FM), fat-free mass (FFM), and total mass, respectively. Custom subregion analysis results in R2s of 0.70-0.89 for left and right thigh composition. We demonstrate the ability of models to produce quantitatively accurate visualizations of soft tissue and bone, confirming a strong relationship between body shape and composition.

CONCLUSIONS

This work highlights the potential of generative models in medical imaging and reinforces the importance of quantitative validation for assessing their clinical utility.

Equations for smartphone prediction of adiposity and appendicular lean mass in youth soccer players

Digital anthropometry by three-dimensional optical imaging systems and smartphones has recently been shown to provide non-invasive, precise, and accurate anthropometric and body composition measurements. To our knowledge, no previous study performed smartphone-based digital anthropometric assessments in young athletes. The aim of this study was to investigate the reproducibly and validity of smartphone-based estimation of anthropometric and body composition parameters in youth soccer players. A convenience sample of 124 male players and 69 female players (median ages of 16.2 and 15.5 years, respectively) was recruited. Measurements of body weight and height, one whole-body Dual-Energy X-ray Absorptiometry (DXA) scan, and acquisition of optical images (performed in duplicate by the Mobile Fit app to obtain two avatars for each player) were performed. The reproducibility analysis showed percent standard error of measurement values < 10% for all anthropometric and body composition measurements, thus indicating high agreement between the measurements obtained for the two avatars. Mobile Fit app overestimated the body fat percentage with respect to DXA (average overestimation of + 3.7% in males and + 4.6% in females), while it underestimated the total lean mass (- 2.6 kg in males and - 2.5 kg in females) and the appendicular lean mass (- 10.5 kg in males and - 5.5 kg in females). Using data of the soccer players, we reparameterized the equations previously proposed to estimate the body fat percentage and the appendicular lean mass and we obtained new equations that can be used in youth athletes for body composition assessment through conventional anthropometrics-based prediction models.

Cross-sectional assessment of body composition and detection of malnutrition risk in participants with low body mass index and eating disorders using 3D optical surface scans

BACKGROUND

New recommendations for the assessment of malnutrition and sarcopenia include body composition, specifically reduced muscle mass. Three-dimensional optical imaging (3DO) is a validated, accessible, and affordable alternative to dual X-ray absorptiometry (DXA).

OBJECTIVE

Identify strengths and weaknesses of 3DO for identification of malnutrition in participants with low body mass index (BMI) and eating disorders.

DESIGN

Participants were enrolled in the cross-sectional Shape Up! Adults and Kids studies of body shape, metabolic risk, and functional assessment and had BMI of <20 kg/m2 in adults or <85% of median BMI (mBMI) in children and adolescents. A subset was referred for eating disorders evaluation. Anthropometrics, scans, strength testing, and questionnaires were completed in clinical research centers. Lin's Concordance Correlation Coefficient (CCC) assessed agreement between 3DO and DXA; multivariate linear regression analysis examined associations between weight history and body composition.

RESULTS

Among 95 participants, mean ± SD BMI was 18.3 ± 1.4 kg/m2 in adult women (N = 56), 19.0 ± 0.6 in men (N = 14), and 84.2% ± 4.1% mBMI in children (N = 25). Concordance was excellent for fat-free mass (FFM, CCC = 0.97) and strong for appendicular lean mass (ALM, CCC = 0.86) and fat mass (FM, CCC = 0.87). By DXA, 80% of adults met the low FFM index criterion for malnutrition, and 44% met low ALM for sarcopenia; 52% of children and adolescents were <-2 z-score for FM. 3DO identified 95% of these cases. In the subset, greater weight loss predicted lower FFM, FM, and ALM by both methods; a greater percentage of weight regained predicted a higher percentage of body fat.

CONCLUSIONS

3DO can accurately estimate body composition in participants with low BMI and identify criteria for malnutrition and sarcopenia. In a subset, 3DO detected changes in body composition expected with weight loss and regain secondary to eating disorders. These findings support the utility of 3DO for body composition assessment in patients with low BMI, including those with eating disorders. This trial was registered at clinicaltrials.gov as NCT03637855.

Accuracy and Precision of 3-dimensional Optical Imaging for Body Composition by Age, BMI, and Ethnicity

BACKGROUND

The obesity epidemic brought a need for accessible methods to monitor body composition, as excess adiposity has been associated with cardiovascular disease, metabolic disorders, and some cancers. Recent 3-dimensional optical (3DO) imaging advancements have provided opportunities for assessing body composition. However, the accuracy and precision of an overall 3DO body composition model in specific subgroups are unknown.

OBJECTIVES

This study aimed to evaluate 3DO's accuracy and precision by subgroups of age, body mass index, and ethnicity.

METHODS

A cross-sectional analysis was performed using data from the Shape Up! Adults study. Each participant received duplicate 3DO and dual-energy X-ray absorptiometry (DXA) scans. 3DO meshes were digitally registered and reposed using Meshcapade. Principal component analysis was performed on 3DO meshes. The resulting principal components estimated DXA whole-body and regional body composition using stepwise forward linear regression with 5-fold cross-validation. Duplicate 3DO and DXA scans were used for test-retest precision. Student's t tests were performed between 3DO and DXA by subgroup to determine significant differences.

RESULTS

Six hundred thirty-four participants (females = 346) had completed the study at the time of the analysis. 3DO total fat mass in the entire sample achieved R2 of 0.94 with root mean squared error (RMSE) of 2.91 kg compared to DXA in females and similarly in males. 3DO total fat mass achieved a % coefficient of variation (RMSE) of 1.76% (0.44 kg), whereas DXA was 0.98% (0.24 kg) in females and similarly in males. There were no mean differences for total fat, fat-free, percent fat, or visceral adipose tissue by age group (P > 0.068). However, there were mean differences for underweight, Asian, and Black females as well as Native Hawaiian or other Pacific Islanders (P < 0.038).

CONCLUSIONS

A single 3DO body composition model produced accurate and precise body composition estimates that can be used on diverse populations. However, adjustments to specific subgroups may be warranted to improve the accuracy in those that had significant differences. This trial was registered at clinicaltrials.gov as NCT03637855 (Shape Up! Adults).

Monitoring body composition change for intervention studies with advancing 3D optical imaging technology in comparison to dual-energy X-ray absorptiometry

BACKGROUND

Recent 3-dimensional optical (3DO) imaging advancements have provided more accessible, affordable, and self-operating opportunities for assessing body composition. 3DO is accurate and precise in clinical measures made by DXA. However, the sensitivity for monitoring body composition change over time with 3DO body shape imaging is unknown.

OBJECTIVES

This study aimed to evaluate the ability of 3DO in monitoring body composition changes across multiple intervention studies.

METHODS

A retrospective analysis was performed using intervention studies on healthy adults that were complimentary to the cross-sectional study, Shape Up! Adults. Each participant received a DXA (Hologic Discovery/A system) and 3DO (Fit3D ProScanner) scan at the baseline and follow-up. 3DO meshes were digitally registered and reposed using Meshcapade to standardize the vertices and pose. Using an established statistical shape model, each 3DO mesh was transformed into principal components, which were used to predict whole-body and regional body composition values using published equations. Body composition changes (follow-up minus the baseline) were compared with those of DXA using a linear regression analysis.

RESULTS

The analysis included 133 participants (45 females) in 6 studies. The mean (SD) length of follow-up was 13 (5) wk (range: 3-23 wk). Agreement between 3DO and DXA (R2) for changes in total FM, total FFM, and appendicular lean mass were 0.86, 0.73, and 0.70, with root mean squared errors (RMSEs) of 1.98 kg, 1.58 kg, and 0.37 kg, in females and 0.75, 0.75, and 0.52 with RMSEs of 2.31 kg, 1.77 kg, and 0.52 kg, in males, respectively. Further adjustment with demographic descriptors improved the 3DO change agreement to changes observed with DXA.

CONCLUSIONS

Compared with DXA, 3DO was highly sensitive in detecting body shape changes over time. The 3DO method was sensitive enough to detect even small changes in body composition during intervention studies. The safety and accessibility of 3DO allows users to self-monitor on a frequent basis throughout interventions. This trial was registered at clinicaltrials.gov as NCT03637855 (Shape Up! Adults; https://clinicaltrials.gov/ct2/show/NCT03637855); NCT03394664 (Macronutrients and Body Fat Accumulation: A Mechanistic Feeding Study; https://clinicaltrials.gov/ct2/show/NCT03394664); NCT03771417 (Resistance Exercise and Low-Intensity Physical Activity Breaks in Sedentary Time to Improve Muscle and Cardiometabolic Health; https://clinicaltrials.gov/ct2/show/NCT03771417); NCT03393195 (Time Restricted Eating on Weight Loss; https://clinicaltrials.gov/ct2/show/NCT03393195), and NCT04120363 (Trial of Testosterone Undecanoate for Optimizing Performance During Military Operations; https://clinicaltrials.gov/ct2/show/NCT04120363).

Machine learning-based obesity classification considering 3D body scanner measurements

Obesity can cause various diseases and is a serious health concern. BMI, which is currently the popular measure for judging obesity, does not accurately classify obesity; it reflects the height and weight but ignores the characteristics of an individual's body type. In order to overcome the limitations of classifying obesity using BMI, we considered 3-dimensional (3D) measurements of the human body. The scope of our study was limited to Korean subjects. In order to expand 3D body scan data clinically, 3D body scans, Dual-energy X-ray absorptiometry, and Bioelectrical Impedance Analysis data was collected pairwise for 160 Korean subjects. A machine learning-based obesity classification framework using 3D body scan data was designed, validated through Accuracy, Recall, Precision, and F1 score, and compared with BMI and BIA. In a test dataset of 40 people, BMI had the following values: Accuracy: 0.529, Recall: 0.472, Precision: 0.458, and F1 score: 0.462, while BIA had the following values: Accuracy: 0.752, Recall: 0.742, Precision: 0.751, and F1 score: 0.739. Our proposed model had the following values: Accuracy: 0.800, Recall: 0.767, Precision: 0.842, and F1 score: 0.792. Thus, our accuracy was higher than BMI as well as BIA. Our model can be used for obesity management through 3D body scans.