By Jimmy Westenberg
Summary
- Wearable devices can assess body fat through bioelectrical impedance analysis, but precise measurements typically necessitate X-ray technology.
- Google Research has developed a new method for providing accurate estimates from selfies.
- This system not only evaluates body composition but may also help in predicting insulin resistance.
A significant number of individuals grapple with weight issues. Research from the NIH indicates that nearly 73.1% of American adults are categorized as overweight or obese. Gaining a comprehensive understanding of body composition extends beyond merely checking weight; the most reliable methods for evaluating fat distribution often involve complex X-ray imaging. However, some researchers at Google are exploring more accessible ways to provide similar insights using smartphone cameras.
PhotoScan is the innovative approach developed by Google Research to estimate body composition by leveraging AI to analyze a series of 2D images. Initially, the system was trained on data derived from actual X-ray scans using dual-energy X-ray absorptiometry (DXA). This technique assesses overall body fat alongside the ratios of fat distribution in different body parts.
By integrating data from MRI scans and refining the model with actual photographs taken from smartphones, the research team has successfully created a model that estimates body fat with improved accuracy compared to traditional methods, such as the bioelectrical impedance analysis (BIA) found in many wearables. This new model also effectively measures A/G and V/S ratios, which are beyond the capabilities of BIA sensors.
In addition to analyzing body composition, Google Research aims to assess whether PhotoScan can aid in predicting conditions like insulin resistance. This focus aligns with Google’s heightened interest in health metrics, as seen in their upcoming health management features on devices like the Pixel Watch.
While the research’s scientific foundation appears robust, the challenge lies in persuading individuals to share potentially unflattering selfies with the application. Even with a strong privacy assurance, this can be a tall order. Furthermore, a critical concern remains: once users receive this data, how can they be motivated to implement meaningful lifestyle and dietary changes?
This represents a challenging hurdle, but innovations like PhotoScan may pave the way for initial progress. At present, Google has not disclosed any intentions to market PhotoScan as a product, yet the technology could become part of future offerings in the realm of Google Health.
