Abstract LB377: Satellite embedding-derived built environment features predict cancer risk factors

Abstract Background: Obesity and diabetes are established modifiable cancer risk factors with strong environmental determinants. The CDC's Social Vulnerability Index (SVI) captures socioeconomic factors associated with cancer risk, but may miss built environment features—walkability, green space, food environment—that independently influence obesogenic behaviors. We evaluated whether satellite-derived embeddings predict cancer risk factors beyond SVI. Methods: We analyzed 68,032 US census tracts using 64-dimensional embeddings from Google DeepMind's AlphaEarth satellite imagery model alongside CDC/ATSDR SVI components. We compared predictive performance for obesity, diabetes, and other cancer risk factors using machine learning with cross-validation, quantifying unique variance explained by embeddings after controlling for SVI. Results: Satellite embeddings improved cancer risk factor prediction beyond SVI alone. For obesity, adding embeddings to SVI increased R² from 0.59 to 0.71; for diabetes, from 0.70 to 0.74; for depression, from 0.44 to 0.56. Embeddings captured 6-11 percentage points of variance unexplained by socioeconomic factors. Embeddings predicted cancer risk factors (obesity R²=0.29, diabetes R²=0.19) but showed limited association with cancer prevalence itself, consistent with the long latency between environmental exposure and cancer diagnosis. Conclusions: Satellite imagery may capture built environment features associated with cancer risk factors that are not fully reflected in socioeconomic indices. These findings suggest potential value in incorporating satellite-derived features into cancer prevention surveillance and warrant further investigation of built environment interventions for reducing cancer risk factor burden. Citation Format: Chris Lim. Satellite embedding-derived built environment features predict cancer risk factors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 2 (Late-Breaking, Clinical Trial, and Invited Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(8_Suppl):Abstract nr LB377.