Seeing beyond BMI: Estimating cardiometabolic risk with smartphone imagery

Google Research has developed PhotoScan, a deep learning tool that estimates body composition metrics from smartphone photos with accuracy comparable to clinical DXA scans, enabling better prediction of insulin resistance and cardiometabolic risk. The approach measures key metrics like body fat percentage, fat distribution ratios, and visceral fat levels—factors that are more predictive of metabolic disease than BMI alone—while offering a non-invasive, accessible alternative to expensive radiation-based clinical scans. This technology could enable earlier detection of metabolic dysfunction before type 2 diabetes and other serious health conditions develop.