Advancing Exposure Assessment of the Built Environment and Studying Its Associations with Various Health Outcomes

The built environment includes physical or manmade features such as streets andtransit networks; green spaces and parks; buildings and other infrastructure, etc. Thismodifiable environment has been increasingly recognized as an important determinantof health in the modern society. When properly designed, the built environment cansignificantly promote the health and well-being of residents. Nevertheless, the builtenvironment involves complicated and interacting infrastructure, with many differentelements included. Traditional exposure assessment of the built environment mosthappened through self-reported questionnaire, or crude metrics from GIS calculationssuch as distance to highway/parks, Normalized Difference Vegetation Index (NDVI), etc.The rapid development of image processing technologies enabled by deep learningmethods has opened the opportunity to advance the exposure assessment of the builtenvironment. It is desirable to apply these methods to analyze the built environment andstudy its effect on a wide spectrum of health outcomes, which can give rise to morerational guidance to urban design and policy recommendations. In this project, deep learning models will be developed on satellite images aroundresidential addresses collected in a database involving about 6,000 middle schoolstudents in a Chinese city. The developed model will be used to segment the builtenvironment into individual features (green space, major roads, buildings, water, etc.).Further, their associations with a set of health outcomes (obesity, asthma and allergicsymptoms, etc) will be investigated. Finally, the overall built environment will be treatedas a composite factor and tested for how much it explains the variation in the healthoutcomes. The developed built environment image segmentation model can be used to processsatellite images and assess built environment exposure for other cities as well. Thefindings of this study can also add value to the current evidence on understanding theimpact of built environment on the health, benefiting public health and urban planningstudies.