3D Optical Scanning for Body Composition: Does it Work?

Picture this: A device that can determine your percent fat from a picture. Does it work?

There is a growing trend where devices are being developed that use either 2 or 3 dimensional photography to determine body composition. This new method is sometimes called digital anthropometry or body scanning. These devices are made by many different manufacturers, are available in different venues, including on smartphones, and use different technology to move from picture to percent fat. There is a wide variety of price points as well, from free phone apps to devices that costs many thousands of dollars. This creates a unique situation, a perfect storm of sorts, where the consumer has increasingly easier access to these methods but who is not necessarily equipped to determine whether the method is accurate. Let’s try to break down the issue and clarify whether these new options are worth using.

First, let’s describe how these devices work. Some devices are 360 degree scanners that literally make a video or a scan of a person. Some devices estimate circumferences of various body areas and then use an equation or algorithm to convert these circumferences and lengths into percent fat. If the system only relies on the circumferences measured, this means that there is an assumption that larger values are due to increased fat, which we know is not an accurate assumption for those who are muscular. Some devices are a bit more sophisticated and use machine learning (or other methods) that enables the device to assess what shapes and circumferences are related to a muscular shape versus a thin shape versus a shape with more fat present. Finally, some devices are programmed to use the camera from your smartphone to take a front view and back view picture of you. These algorithms used in these devices were developed from data on people who were measured with an accurate method and then “trained” to recognize what shape indicates fatness and which indicates muscularity.

Do they work? Most show errors that are no better from established field methods. I will highlight 3 papers (Tinsley, 2020; Cabre, 2021; Majmudar, 2022). In 2020, Tinsley tested 4 commercially available devices while Cabre in 2021 studied one of the same devices tested by Tinsley. The 4 devices all showed errors higher than what we consider to be acceptable for field methods. The most accurate device was the 3D Body Scanner by Fit3D (San Mateo, California) with an overall error of predicting an individual’s percent fat of plus or minus 3.7%. However, this device significantly overestimated percent fat in those who had less fat and underestimated percent fat in those with more fat. The device that was tested by both Tinsley and Cabre was the Styku scanner (Los Angeles, CA). In both studies, the device overestimated percent fat in those who had less fat and underestimated percent fat in those with more fat. In addition, the percent fat of the group of people measured was systematically overestimated (by 4%) in the Cabre study and underestimated (by 4%) in the Tinsley study. Completely opposite results! However, in both studies, the individual prediction error was too high (plus or minus 5-6%).

In 2022, in a report by Majmudar, a smartphone (app) based 2-dimensional system that utilized 2 photos (front and back) was tested on 138 people. They developed the algorithm from body composition measured by DXA. In this sample, the app performed well in a diverse group of people with a range of BMI’s but there was significant overestimation in those with more fatness (and underestimation in those with low fatness). However, there was no cross validation on a new sample of people and the Standard Error of Estimate was not reported. These are critical aspects that are needed for a full evaluation of a new method.

The difficulty with any body composition method is that it’s possible for them to show good accuracy in one study but that doesn’t mean the method should be considered accurate in general. Thorough testing and cross validation are needed, along with good agreement with an established method, before any method can be recommended for use. The bottom line for 3D optical scanning and smartphone apps for estimating body composition is that we need more testing on large diverse samples of people AND low errors need to be consistently shown. Until that time, we remain skeptical of this technology.

References:

Cabre H et al (2021) Validity of a 3-dimensional body scanner: comparison against a 4-compartment model and dual energy X-ray absorptiometry. Appl. Physiol. Nutr. Metab. 46: 644–650.

Tinsley G et al. (2020) 3-Dimensional optical scanning for body composition assessment: A 4-component model comparison of four commercially available scanners. Clin. Nutr. 39: 3160-3167.

Majmudar M et al. (2022) Smartphone camera based assessment of adiposity: a validation study. npj Dig. Med. 5:79