Google Develops Photo-Based AI to Predict Insulin Resistance

Insulin resistance is a key warning sign for type 2 diabetes and other metabolic disorders, but detailed body-fat assessment often requires specialized clinical equipment. Google’s PhotoScan deep-learning framework aims to estimate body composition from smartphone photographs, potentially offering a more accessible and non-invasive way to identify people who may face elevated metabolic risk before symptoms emerge.

Google’s latest disclosed research shows PhotoScan analyzing smartphone images to estimate fat composition and predict insulin-resistance risk, with accuracy approaching dual-energy X-ray absorptiometry, or DXA, the clinical gold standard. The technology remains under development and would require validation across larger and more diverse populations before routine medical use, but it could eventually broaden early screening beyond hospitals and specialist clinics.