These proof-of-concept observations, the first of their kind, indicate that smartphone applications such as MeasureNet can now fill the void in WHR measurements made in clinical and home settings.

The smartphone approach can potentially displace 3D scanning methods 18 that are more costly and impractical to implement outside of specialized research and clinical facilities. Human shape and pose estimation are active areas of research in the computer vision and machine learning CVML communities.

Most of the current approaches predict body shape using a learned model or fit body shape using an optimization-based approach with SMPL 19 , a parametric 3D body model given observations such as 2D key points, silhouettes, or images 20 , 21 , 22 , 23 , Recent developments as reported by Sengupta et al.

In contrast to their approach, we focused directly on estimating body circumferences and derived measures such as the WHR, a strategy we found more accurate than estimating body circumferences from the reconstructed body model.

Our MeasureNet model estimates circumferences and the WHR directly, and uses SMPL, the parametric 3D body model only as a regularizer during training.

This allowed the circumference predictions to be independent of the space of SMPL parameters. Several challenges needed to be overcome on the path to developing MeasureNet. First, MeasureNet needed to generalize to different body shapes and be invariant to lighting and background conditions, clothes worn, user distance from the smartphone, and smartphone type.

Our MeasureNet algorithms account for all of these factors and conditions that became apparent during the software development phase. Another factor posing a development challenge was that training accurate CVML models required access to accurate ground truth measurements.

Manual measurements of waist and hip circumferences, however, tend to be error prone as reported by Sebo et al. On the other hand, using highly accurate 3D laser scanners to extract ground truth measurements is expensive and time consuming.

We addressed both problems by training a CNN on realistic-looking synthetic data sampled according to an empirical distribution, and we demonstrated strong generalization high accuracy and repeatability to real, previously unseen test images.

Adding WHR estimates to clinical and self-evaluations improves health risk predictions beyond those of BMI and other currently available biometrics Larger visceral adipose tissue volumes and waist circumferences are associated with greater risks of adverse health outcomes 7 , 8 , 9 , By contrast, larger subcutaneous gluteofemoral adipose tissue volumes and hip circumferences are associated with a reduced risk of developing multiple cardiovascular and metabolic outcomes 6.

WHR or the individual waist and hip circumferences can also be added to health outcome prediction models now in development by our group and others. Large-scale studies designed to identify health-risk genetic markers can use programs like MeasureNet to accurately capture participant shape using their own smartphones.

Anticipated camera advancements and future machine learning algorithm refinements over time will further expand the applicability of smartphone phenotyping methods.

There are several limitations with our developed model that form the potential basis of future research. As part of the realistic sampling process the current SMPL 3D mesh model was estimated using 3D scans covering the US general population and therefore is biased towards the average North American population.

This kind of potential bias can be removed by including 3D scans of participants outside of the US when estimating the SMPL 3D mesh model. A subset of participants In the CSD dataset had only one measurement taken by trained clinical staff. Therefore, the resulting ground truth measurement can be noisy and it can affect the accuracy metrics.

The MeasureNet model is trained using synthetic training data. However, the current synthetic data generator can only represent the shape and pose of a minimally clothed body and fails to model complex topology of loose clothing. This results in a synthetic-to-real domain gap that reduces the accuracy of MeasureNet.

A more realistic synthetic data generator that can model loose clothing can help alleviate this issue. Further studies to understand the relationship between MeasureNet and health risks can help determine the desired accuracy level needed for an accurate health risk prediction.

Addressing this bias requires the inclusion of participants from underrepresented groups to foster a more balanced and equitable dataset. In conclusion, the current study fills a long-held gap in accurately and reproducibly quantifying the WHR, an extensively researched health-risk biometric, outside of specialized facilities.

The developed novel software, MeasureNet, can operate on conventional smartphones and thus vastly extend shape phenotyping capabilities to a large percentage of the global population, even to remote settings. Future studies are needed to extend software capabilities to populations beyond those in North America and to non-adult age groups.

The study hypothesis was tested in two phases. A smartphone application based on computer vision algorithms was developed in the first study phase. The development of this algorithm, MeasureNet, is described in the methods section that follows. The second phase involved testing MeasureNet performance in a series of experimental studies Supplementary Fig.

First, the accuracy of MeasureNet and self-measurements were compared to flexible tape measurements taken by trained staff in a sample of healthy adults referred to as the Circumference Study Dataset CSD. Accuracy metrics are defined in the Statistical Methods section.

Circumferences were measured according to NHANES guidelines Supplementary Note 1. MeasureNet and self-measurements were compared to the ground truth tape measurements. A second experimental study involved comparison of MeasureNet to state-of-the-art approaches for three-dimensional 3D shape estimation.

Specifically, we compared MeasureNet, SPIN 20 , STRAPS 21 , and recent work by Sengupta et al. This dataset is referred to as the Human Solutions dataset. We had front-, side-, and back-viewpoint color images, height, and body weight for each participant along with their 3D laser scan.

The Skinned Multi-Person Linear SMPL model was fit to each 3D scan to estimate the shape and pose of the scan We extracted the ground truth circumferences from the fitted SMPL model at predefined locations corresponding to hip, waist, chest, thigh, calf and bicep as shown in Supplementary Figs.

Third, we measured the noise in tape measurements compared to MeasureNet using data from a subset of healthy men and women evaluated in the CSD dataset. Each person was measured twice by a trained staff member staff measurements and two sets of images were also taken by the staff member MeasureNet.

Each person also measured themselves twice using measuring tape self-measurements. For staff measurements, each person was measured by two different staff members to ensure minimal correlation between consecutive measurements. We used the difference between two consecutive measurements to analyze the noise distributions of staff-measurements, MeasureNet, and self-measurements.

Lastly, we compared accuracy and repeatability of our approach to the ground truth on a synthetic dataset. We created the dataset by rendering each synthetically generated mesh using different camera parameters height, depth, focal length and different body poses placed in front of randomly selected backgrounds.

The dataset was generated using synthetic meshes of men and women. This data is referred to as the Synthetic Dataset. We considered all of the renderings for a particular mesh to measure repeatability robustness of our approach. Repeatability metrics are defined in the Statistical Methods section.

Different factors such as background, camera parameters, and body pose changes were present across multiple renderings of the same mesh. A repeatable approach should ideally predict the same output for different renderings of the same mesh. We also use this dataset to evaluate accuracy given all of the renderings and their ground truth.

A flow diagram showing the multiple study human participant evaluations is presented in Supplementary Fig. Consent was obtained for the collection and use of the personal data voluntarily provided by the participants during the study.

An overview of our approach for measuring WHR is shown in Fig. The user inputs their height, weight, and sex into their smartphone. Voice commands from the application then guide the person to capture front-, side-, and back-viewpoint color images. The images are then automatically segmented into 23 regions such as the background, upper left leg, lower right arm, and abdomen by a specialized convolutional neural network CNN trained to perform semantic image segmentation.

Intuitively, this step suppresses irrelevant background features, provides additional spatial context for body parts, and affords important benefits during model training, which we will discuss subsequently. Overview of the anthropometric body dimension measurement approach.

The user first enters their height, weight, and sex into the smartphone application. Voice commands then position the user for capture of front, side, and back color images. The images are then segmented into semantic regions using a segmentation network.

The segmentation results are then passed to a second network referred to as MeasureNet that predicts WHR and body circumferences. Each input is passed through a modified Resnet network which is then concatenated and passed through Resnet-4, self-attention block and a fully connected layer FC layer before predicting WHR and body circumferences.

Resnet is modified to include Squeeze-Excitation blocks SE. CNN, convoluted neural network. Synthetic images are used to train this model.

Real images are used during inference after the model is trained. Color images shown in the figure are synthetically generated. Features from each view are then concatenated together and fed to a Resnet-4 network and a self-attention network 28 followed by a fully connected layer to predict body circumferences and WHR as illustrated in Fig.

Direct prediction of circumferences: Predicting body circumferences directly outperformed first reconstructing the body model 3D SMPL mesh 19 and then extracting measurements from it. Number of input views: Using three views of the user as input improved the accuracy as compared to using one or two views of the user.

Tables 2 and 3 shows the improvement in accuracy with increasing number of input views and using direct prediction of circumferences. Swish vs. ReLU activations: Resnet typically uses ReLU activations Self-Attention and Squeeze-Excitation for non-local interactions: Including squeeze-excitation blocks 27 with Resnet branches for cross-channel attention and a self-attention block 28 after the Resnet-4 block allowed the model to learn non-local interactions e.

Supplementary Note 4 shows the accuracy improvements due to self-attention, squeeze-excitation and Swish activation blocks. Sex-specific model: Training separate, sex-specific MeasureNet models further improved accuracy. As we show in Tables 2 and 3 , sex-specific models have lower prediction errors compared to sex-neutral models.

MeasureNet predicts multiple outputs, such as body shape, pose, camera, volume, and 3D joints. Predicting multiple outputs in this way multi-tasking has been shown to improve accuracy for human-centric computer vision models Additionally, MeasureNet predicts circumferences and WHR.

Some of the outputs e. The inputs and outputs to MeasureNet are shown in Fig. Important MeasureNet outputs related to circumferences and WHR are:.

Dense Measurements: MeasureNet predicts circumferences defined densely over the body. Details are presented in Supplementary Note 2. Dense measurements reduce the output domain gap between synthetic and ground truth by finding the circumference ring out of circumference rings that minimize the error between tape measurements taken by trained staff and synthetic measurements at a particular ring.

The table in Supplementary Note 2 shows that the predicted error at the optimal circumference ring is the lowest and therefore it is well-aligned with the staff measurements. WHR Prediction: Our model can predict WHR both indirectly by taking ratios of waist and hip estimates and directly i. WHR related outputs are shown in Fig.

The final WHR prediction is an ensemble result, i. As shown in Supplementary Note 5 , we found that the ensemble prediction had the lowest repeatability error most robust without losing accuracy as compared to individual predictions via regression, classification or taking the ratio of waist and hip.

We include training losses on shape, pose, camera, 3D joints, mesh volume, circumferences and waist-hip ratio through classification and regression. The losses are defined in Supplementary Note 6. Since we have multiple loss functions, hand-tuning each loss weight is expensive and fragile.

Based on Kendall et al. Supplementary Note 7 shows the improvement in accuracy when using uncertainty-based loss weighting during training.

MeasureNet was trained with synthetic data. Using synthetic data helps avoid expensive, manual data collection and annotation. However, it comes at the cost of synthetic-to-real domain gap, which leads to a drop in accuracy between a model trained with synthetic data but tested on real data.

We reduced the domain gap by simulating a realistic image capture process on realistic 3D bodies with lifelike appearance texture. Examples of synthesized body shapes for different BMI values are shown in Fig.

The SMPL mesh model 19 is parameterized by shape and pose parameters. To encourage realism in the synthetic dataset and minimize domain gap, it was important to sample only realistic parameters and to match the underlying distribution of body shapes of the target population.

Our sampling process was used to generate approximately one million 3D body shapes with ground truth measurements, and consisted of three steps:. Fit SMPL parameters: Given an initial set of 3D scans by a laser scanner as a bootstrapping dataset, we first fitted the SMPL model to all scans 19 to establish a consistent topology across bodies and to convert each 3D shape into a low-dimensional parametric representation.

Due to the high fidelity of this dataset and the variation across participants, we used this dataset as a proxy for the North American demographic distribution of body shapes and poses. Cluster samples: We recorded the sex and weight of each scanned subject, and extracted a small set of measurements from the scan, such as height, and hip, waist, chest, thigh, and bicep circumferences.

We trained a sex-specific Gaussian Mixture Model GMM to categorize the measurements into 4 clusters we found the optimal number of clusters using Bayesian information criterion.

Sample the clusters using importance sampling: Finally, we used importance sampling to match the likelihood of sampling a scan to match the distribution across all clusters. This allowed us to create a large synthetic dataset of shape and pose parameters whose underlying distribution matched the diversity of the North American population.

NHANES was collected by the Center for Disease Control and Prevention between the years and and consists of the demographics, body composition and medical conditions of about , unique participants from North American population. Valid renderings are images in which body shapes are visible from at least the top of the head to the knees.

This ensures that the sampled camera parameters match the realistic distribution of camera parameters observed for real users. An example of realistic sampling of shape, pose, and camera are shown in Fig. Example of realistic sampling of body shape, pose, and camera simulating the image capture process.

Once body shape, body pose, and camera orientation were sampled, we transferred the texture from a real person onto the 3D mesh, placed it in front of a randomly selected background image of an indoor scene and rendered a realistic color rendering given the camera pose.

The textured and realistic color rendering was then segmented using the segmentation network that was used as an input to train MeasureNet. The ground truth targets used to train MeasureNet were extracted from sampled synthetic mesh. Transferring the texture from a real person allowed us to generate diverse and realistic samples and had two main advantages.

First, we transferred the texture from a real person which avoided manually generating realistic and diverse textures. Through this method, we generated a texture library of forty thousand samples using trial users different from test-time users.

Second, since we segmented the color images using a trained segmentation model, we did not have to include additional segmentation noise augmentation 30 during training.

This is in contrast to the existing methods 21 , 30 that add segmentation noise to the synthetic image in order to simulate the noisy segmentation output during test-time. We used the segmented image as input to MeasureNet instead of a textured color image to force MeasureNet to not use any lighting or background-related information from the synthetic training data which can have different distributions during training and testing.

In Supplementary Note 8 , we show that training a model with textured color image generalizes poorly when tested on real examples as compared to segmented images. Intuitively, we believe this is the case because synthetic textured color images lack realism on their own, but generate realistic segmentation results when passed through a semantic segmentation model.

Overall, the texture transfer process consisted of two steps. First, we created a texture library by extracting textures from real images using our participant pipeline. We extracted around forty thousand texture images from trial users. Second, given the texture images, we rendered a randomly sampled synthetic mesh using a random texture image, rendered it on a random background, and passed it through the segmentation.

The process of realistic textured rendering by transferring the texture from a real person synthetic in this case is shown in Fig. The renderings when segmented using fixed segmentation network were used as input to train MeasureNet.

The end-to-end training process for MeasureNet is shown in Fig. The ground truth targets used to train MeasureNet are extracted from sampled synthetic mesh.

Generation of realistic color mesh renderings by transferring texture from a real person synthetic in this example. The renderings when segmented using a fixed network are used as input to train MeasureNet. Training of MeasureNet model using realistic synthetic data.

Given a sampled synthetic mesh, realistic synthetic images are generated that are segmented. The segmented images are used as input to MeasureNet and corresponding predictions are compared against the ground truth extracted from synthetic mesh. The accuracy of MeasureNet and self-measurements were compared to trained staff-measured ground truth estimates in the CSD using mean absolute error MAE; Eq.

MAPE is similar to MAE but calculates mean relative percentage error. G i is the ground truth, P i is the prediction, and n is the number of users. MAE was also used for comparing MeasureNet to other state-of-the-art approaches for estimating circumferences and WHR.

The same procedures were used for evaluating noise in staff measurements, MeasureNet predictions, and self-measurements. Noise was estimated by plotting histograms of the between-measurement or prediction differences meas 1 and meas 2.

Biases in differences were removed before plotting the histograms by including the Δs in both directions: meas 1 — meas 2 and meas 2 — meas 1. We also fit Gaussian curves on the resulting histograms to estimate the noise standard deviations. Repeatability was computed on the synthetic dataset and measured as the mean and 90 th percentile P90 of absolute differences.

The participant data evaluated in this study is approved by PBRC Institutional Review Boards clinicaltrials. gov identifier: NCT The reported investigation extends the analyses to anthropomorphic data waist and hip circumference measurements , and reflects a secondary analysis of data collected by Amazon vendors in commercial settings.

All participants signed consents in these no-risk studies that granted full permission to use their anonymized data. The investigators will share the data in this study with outside investigators upon request to and approval by the lead author.

Further information on research design is available in the Nature Research Reporting Summary linked to this article. The data that supports the findings of this study is available from the corresponding author upon reasonable request and approval of Amazon Ltd. The code that supports the findings of this study is available from the corresponding author upon reasonable request and approval of Amazon Ltd.

Custom scripts for data processing were developed in MATLAB a and statistical analyses were performed in Python version 3. Vague, J. Sexual differentiation; Factor determining forms of obesity. Presse Med. CAS Google Scholar. Krotkiewski, M. Impact of obesity on metabolism in men and women.

Importance of regional adipose tissue distribution. Article CAS PubMed PubMed Central Google Scholar. Larsson, B. et al. Abdominal adipose tissue distribution, obesity, and risk of cardiovascular disease and death: 13 year follow up of participants in the study of men born in Article CAS Google Scholar.

Bjorntorp, P. Fat cell distribution and metabolism. Y Acad. Article CAS PubMed Google Scholar. World Health Organization WHO. Waist circumference and waist-hip ratio: report of a WHO expert consultation World Health Organization, Cameron, A.

A systematic review of the impact of including both waist and hip circumference in risk models for cardiovascular diseases, diabetes and mortality. Cerhan, J. A pooled analysis of waist circumference and mortality in , adults. Mayo Clin. Article PubMed Google Scholar. Jacobs, E.

Waist circumference and all-cause mortality in a large US cohort. Intern Med. Ross, R. Waist circumference as a vital sign in clinical practice: a Consensus Statement from the IAS and ICCR Working Group on Visceral Obesity. Article PubMed PubMed Central Google Scholar.

Seidell, J. Criminisi, A. Normalized sensitivity of multi-dimensional body composition biomarkers for risk change prediction. Sebo, P. Swiss Med. Wkly , w PubMed Google Scholar. Accuracy of anthropometric measurements by general practitioners in overweight and obese patients.

Similar results were obtained for hip circumference. Conclusion: Considerable variation in waist and hip circumferences and WHR were observed among the study populations. Waist circumference and WHR, both of which are used as indicators of abdominal obesity, seem to measure different aspects of the human body: waist circumference reflects mainly the degree of overweight whereas WHR does not.

Abstract Objective: To assess differences in waist and hip circumferences and waist-to-hip ratio WHR measured using a standard protocol among populations with different prevalences of overweight. Publication types Multicenter Study Research Support, Non-U.

Gov't Research Support, U.

Waist circumference and all-cause mortality in a large US cohort. Arch Intern Med. Jensen MD, Ryan DH, Apovian CM, et al. Moyer VA. Screening for and management of obesity in adults: U. Preventive Services Task Force recommendation statement.

Ann Intern Med. By Richard N. Fogoros, MD Richard N. Fogoros, MD, is a retired professor of medicine and board-certified in internal medicine, clinical cardiology, and clinical electrophysiology. Use limited data to select advertising.

Create profiles for personalised advertising. Use profiles to select personalised advertising. Create profiles to personalise content.

Use profiles to select personalised content. Measure advertising performance. Measure content performance. Understand audiences through statistics or combinations of data from different sources.

Develop and improve services. Use limited data to select content. List of Partners vendors. Heart Health. Heart Disease. Fogoros, MD. Medically reviewed by Anisha Shah, MD. Waist-to-Hip Ratio Scores Men 1. Verywell Health uses only high-quality sources, including peer-reviewed studies, to support the facts within our articles.

Read our editorial process to learn more about how we fact-check and keep our content accurate, reliable, and trustworthy. Fat is more buoyant less dense than water, so someone with high body fat will have a lower body density than someone with low body fat.

This method is typically only used in a research setting. This method uses a similar principle to underwater weighing but can be done in the air instead of in water. Individuals drink isotope-labeled water and give body fluid samples.

Researchers analyze these samples for isotope levels, which are then used to calculate total body water, fat-free body mass, and in turn, body fat mass.

X-ray beams pass through different body tissues at different rates. So DEXA uses two low-level X-ray beams to develop estimates of fat-free mass, fat mass, and bone mineral density. These two imaging techniques are now considered to be the most accurate methods for measuring tissue, organ, and whole-body fat mass as well as lean muscle mass and bone mass.

Measurements of Adiposity and Body Composition. In: Hu F, ed. Obesity Epidemiology. Press Med. Ohlson LO, Larsson B, Svardsudd K, et al. The influence of body fat distribution on the incidence of diabetes mellitus. Larsson B, Svardsudd K, Welin L, Wilhelmsen L, Bjorntorp P, Tibblin G.

Abdominal adipose tissue distribution, obesity, and risk of cardiovascular disease and death: 13 year follow up of participants in the study of men born in Br Med J Clin Res Ed. Zhang C, Rexrode KM, van Dam RM, Li TY, Hu FB. Abdominal obesity and the risk of all-cause, cardiovascular, and cancer mortality: sixteen years of follow-up in US women.

Zhang X, Shu XO, Yang G, et al. Abdominal adiposity and mortality in Chinese women. Arch Intern Med. Despres JP. Health consequences of visceral obesity. Ann Med.

de Koning L, Merchant AT, Pogue J, Anand SS. Waist circumference and waist-to-hip ratio as predictors of cardiovascular events: meta-regression analysis of prospective studies. Heart J.