PI: Perry Hystad
Project number: R21ES029722
Project dates: 8/17/2019-7/31/2021
The built environment is an important modifiable determinant of human health, yet our ability to understand its effects on human health have been limited by the lack of scalable data on specific components (and exposures) of the built environment. The emergence of ubiquitous geo-referenced imagery in the United States (e.g. Google Street View Imagery), combined with recent advances in image processing using deep learning algorithms, offers unprecedented opportunity for measuring street-level built environment features at scales needed for population-based research. To develop and demonstrate the potential of deep learning algorithms for environmental health research we will: develop methods to assess green space features using street view imagery and deep learning algorithms; create new deep learning algorithms to predict urban green space quality, stress reduction and restorative potential; and apply new street view measures to 9,070 adult Twin Pairs in the Washington Twin Registry to determine associations between green space and mental health. Our proposed study will dramatically move the field of environmental health forward by provided a completely new, transferable and scalable exposure assessment method for assessing built environment exposures relevant to human health and provide robust information on how urban green space influences mental health. Overall, our new approach will provide rich new data sources for environmental epidemiologists, city planners, policy makers and neighborhoods and communities at large.