Google Research is sharing details about PhotoScan AI that can estimate body fat using just smartphone photos. Getting a full picture of your body composition requires more than just stepping on a scale. It typically requires complex imaging such as X-ray analysis to assess how and where the body stores fat. Google researchers have been working on a much simpler way to get similar insight using just the phone’s camera.
Google’s PhotoScan AI aims to offer a simpler way to track body composition
The new PhotoScan system estimates body composition by using AI to analyze a series of 2D images of your body. Google apparently trained the system using data from DXA scans, measuring both overall body fat levels and ratios between core and waist/leg fat (A/G ratio) and between organ and subcutaneous fat (V/S ratio).
Combining that with body data from MRI scans and then further refined with a fresh set incorporating actual photos from smartphone cameras, the team’s model was apparently able to visually estimate body fat with better accuracy than with tools like the bioelectrical impedance analysis (BIA) sensor on wearables. It also showed high accuracy in estimating those A/G and V/S ratios, which BIA sensors can’t.
Moving beyond body composition, Google Research also explored whether PhotoScan could help predict conditions like insulin resistance. That seems to be an area Google has been focusing on lately. The researchers found that adding PhotoScan data improved the accuracy of their predictions, with results nearly matching those from clinical DXA scans. In comparison, smartwatch BIA data did not improve the predictions.
More accessible way
Overall, Google Research says its PhotoScan technology could offer a more accessible way to estimate body composition using a smartphone. Clinical DXA imaging delivers the most accurate body composition but lacks scalability. Meanwhile, BIA sensors on wearables offer convenience but are pretty basic. Google says that the PhotoScan approach offers a middle ground. It can estimate granular body composition from standard smartphone imagery with near-DXA accuracy.
While the technology is still a research prototype, the company says it could eventually help with non-invasive screening for insulin resistance and other cardiometabolic health risks. Google also intends to combine PhotoScan with wearable data, glucose readings, and other health information for a more complete picture of metabolic health.
The post Google’s New AI Tool Can Estimate Body Fat Using Smartphone Photos appeared first on Android Headlines.
