Using Convolutional Neural Networks to Derive Neighborhood Built Environments from Google Street View Images and Examine Their Associations with Health Outcomes

Key takeaways:

  • Google Street View (GSV) was used to collect, investigate, and characterize images of neighborhood built environments. Computer vision was used to analyze the images and produce key indicators (characteristics of the built environment). The indicators were then examined for their potential influence on chronic disease and health behaviors.
  • Indicators of urban development (two or more cars, streetlights, street signs) and walkability (crosswalks, sidewalks, presence of apartment and commercial buildings) were associated with lower chronic health conditions including obesity, high blood pressure, high cholesterol, diabetes, smoking, poor mental health, and depression.
  • Single-lane roads were used as an indicator of lower levels of urban development and correlated with increased chronic conditions in rural areas.
  • Chain-link fences were used as an indicator of physical disorder and correlated with poor mental health and depression but also decreased smoking and improved sleep.

Implications:

  • Using GSV, researchers could specify neighborhood features of interest to conduct a sweeping review of neighborhoods across the United States from remote locations. Broadly, this data is beneficial for policy research, change, and implementation. More narrowly, this study shows that GSV can complement data collected during a walk audit.  Nothing can replace feet on the ground, walking the streets. However, GSV can be used as a supplement or to extend the boundaries of a study area.
  • This study contributes to the literature connecting physical activity and mental health. Neighborhood walkability contributes to increased social interactions, reduced anxiety, and lower levels of depression. The ability to walk in your neighborhood creates the opportunity to connect with neighbors and build relationships.
  • Changes to the built environment, such as new and/or improved sidewalks can lead to increased physical activity and improved population health. The health impacts are amplified when walking or rolling is a convenient, easy, and enjoyable option to get around one’s neighborhood or community.
  • Public health agencies and local government should work together to increase neighborhood walkability and create an environment conducive to health. Working in coordination maximizes the impact of community investments.

Yue, X., Antonietti, A., Alirezaei, M., Tasdizen, T., Li, D., Nguyen, L., Mane, H., Sun, A., Hu, M., Whitaker, R.T., and Nguyen, D. “Using Convolutional Neural Networks to Derive Neighborhood Built Environments from Google Street View Images and Examine their Associations with Health Outcomes.” International Journal of Environmental Research and Public Health, 19 (2022).

article cover
filed under