[CLSTR]
< Back to situations

Monitor this situation.

[SITUATION] · [QUIET] · [TECHNOLOGY]

2 clusters · 14 sources · 6 days · First seen · Last updated

Google PhotoScan AI body fat estimation development

Overview

Google Research has developed an experimental AI framework named ‘PhotoScan’ that estimates body composition through smartphone photographs. By analyzing frontal and side-view images alongside height and weight, the deep learning model can estimate body fat percentage, waist-to-hip ratios, and the relationship between visceral and subcutaneous fat.

Research indicates the system shows strong concordance with medical-grade Dual-energy X-ray Absorptiometry (DXA) scans and can predict insulin resistance with accuracy approaching clinical standards. In comparative testing, the technology outperformed bioelectrical impedance analysis (BIA) sensors found in smartwatches and smart scales, averaging within 2.15 percentage points of medical reference values compared to 2.91 percentage points for BIA.

The model was trained using over 35,000 records from the UK Biobank, utilizing MRI data to create 2D projections. While the technology offers an accessible way to monitor metabolic health and fat distribution—factors considered more critical to health than Body Mass Index (BMI)—it remains a research prototype and is not yet a validated medical diagnostic tool.

Entities

Google · UK Biobank · Garmin · Pixel Watch 5 · Gemini

Timeline

  1. 17 days ago

    [TECHNOLOGY] 2 sources
    Google develops AI to estimate body fat via smartphone photos

    Google is developing ‘PhotoScan’, an AI system that uses smartphone photos to estimate body fat and distribution, outperforming bioimpedance analysis in accuracy tests.

  2. 22 days ago

    [TECHNOLOGY] 13 sources
    Google develops AI tool to estimate body fat via smartphone photos

    Google Research is developing PhotoScan, an AI tool that uses smartphone photos to estimate body fat and insulin resistance with accuracy levels comparable to medical DXA scans.

Sources

ad-hoc-news.de · assodigitale.it · es.digitaltrends.com · futurezone.at · imtest.de · ipaddisti.it · it-boltwise.de · manualdohomemmoderno.com.br · netthings.pt · primorska.info · scinexx.de · tech.everyeye.it · tek.sapo.pt · xataka.com.mx