Overview
- This week Google Research published details of PhotoScan, an investigational AI that maps ordinary smartphone images to clinical body‑composition metrics trained from DXA and MRI data.
- In internal tests PhotoScan produced body‑fat estimates within about two percentage points of DXA and outperformed smartwatch bioelectrical impedance analysis at several measures.
- PhotoScan can estimate granular ratios such as android/gynoid and visceral/subcutaneous fat that BIA sensors cannot, and adding PhotoScan outputs improved models for predicting insulin resistance nearly to DXA levels.
- Google says PhotoScan is a research prototype not yet available to consumers and that it plans further validation, broader testing across diverse bodies and work to combine camera outputs with wearable and glucose data.
- If deployed, the tool could make clinical‑grade body‑composition screening far more accessible, raise questions about privacy and regulatory oversight, and create competitive pressure on devices that rely on BIA measurements.