Simple phone images can be converted into remarkably precise body fat measurements using Google’s PhotoScan AI.
TL;DR
- Wearables can estimate body fat using bioelectrical impedance analysis, but really accurate readings require X-ray scans.
- Google Research shares a new technique for producing high-quality predictions based on selfies.
- Beyond only breaking down body composition, the algorithm might potentially be used to predict insulin resistance
The majority of us struggle with our weight. Nearly three out of four American adults (73.1%) are overweight or obese, according to the NIH. However, gaining a whole picture of your body composition is far more difficult than simply using a scale, and some of the most precise methods for determining how and where we carry fat depend on intricate X-ray analysis. Using just the cameras on our phones, some Google researchers have been developing a simpler method to provide the same kind of knowledge.
is a technology developed by Google Research that uses artificial intelligence (AI) to examine a number of 2D pictures of your body in order to estimate body composition. Initially, Google trained the system using some of the data from those real X-ray scans (a method known as dual-energy X-ray absorptiometry, or DXA), which measured body fat levels overall as well as ratios between organ and subcutaneous fat (V/S ratio) and core and waist/leg fat (A/G ratio)
The team’s final model showed the ability to visually estimate body fat with a higher degree of accuracy than the measurements taken with tools like the bioelectrical impedance analysis (BIA) sensor on wearables after combining that with body data from MRI scans and then refining it with a new set that included actual photos from smartphone cameras. Furthermore, it demonstrated excellent accuracy in calculating some A/G and V/S ratios, which BIA sensors are unable to even try.
Google Research sought to see whether PhotoScan results might be used to anticipate disorders such as insulin resistance, in addition to trying to provide insight into body composition. As evidenced by the impending Insulin Resistance Trends in Health Guardian on the Pixel Watch, Google has recently made this a very apparent focus.
It is one thing to proclaim that you have a non-invasive, cost-effective new method of providing people with some of these helpful body composition data points, even though the study appears to be rather sound. Even with the strongest privacy policy in the world, you still need to persuade people to share a lot of potentially embarrassing photographs with the PhotoScan feature. The largest issue of all is still present: How can you truly persuade users to undertake the significant dietary and lifestyle adjustments they require, even after presenting them with this data?
That is a completely separate and much more challenging problem, but maybe tools like PhotoScan will eventually make it easier to at least put us in a position to begin tackling it. We would not be shocked to see this technology appear as a future Google Health solution, but Google is not currently disclosing any plans to make PhotoScan a real product.
