AI-Enhanced Analysis of Built Environment Imagery and Neighborhood Obesity in US Cities
An erratum to this article was published (JAMA Netw Open, 1 Dec 2025 ↗)
Authors' conclusion
In this study, integrating AI-derived built environment features from geospatial imagery was associated with enhanced ability to explain and estimate neighborhood-level obesity prevalence beyond conventional DSE and SDOH factors.
Study context
- Population
- US Census tracts within 94 of the 100 most populous US cities; 2023 CDC PLACES and 2019 American Community Survey data linked to satellite and Street View imagery
- Participants
- 14,413 census tracts
- Exposure
- Model with vs without artificial intelligence (AI) image-derived built environment features
- Outcome
- Crude obesity prevalence at census tract level (adults aged ≥18 years with BMI ≥30.0)
Methodological detailsAI-enhanced modeling framework, census tract as unit of analysis, linear mixed-effects model, 94 498 satellite and 670 860 street view images processed using convolutional neural networks
- AI-enhanced modeling framework
- Census tract as unit of analysis
- Linear mixed-effects model
- 94 498 satellite and 670 860 street view images processed using convolutional neural networks
Main finding
As reported in the abstract
Full model (demographic and socioeconomic (DSE), social determinants of health (SDOH), satellite and street view features), conditional R2 (including random effects)
0.926
Marginal R2 increase on adding image-derived features to DSE and SDOH model
0.632 to 0.745
χ2 = 1303.4P < .001
Observational data: associations, not effects.
Further findings
Tract obesity prevalence
- Median (IQR) obesity prevalence
- 32.4% (26.6%-38.8%)
Original publication
Composed by AI from the published abstract and checked against it. How Briefs are made →