Application of deep learning to heterogeneous open data for measuring urban environment and health
Lead Research Organisation:
Imperial College London
Department Name: School of Public Health
Abstract
In today's world, there is a great deal of information from sources like Google Street View and mobile phone usages about our environment and movement. The promise of big data, in many research fields, business, policy-making alike, that it will transform and improve the way we do things. Researchers in environmental health are also increasingly interested to use of information generated by mobile phones, satellites, wearable devices, social media, etc. The main challenge, however, is that making sense of diverse data at such big scales is not easy. Further, these new datasets are characteristically very different from conventional survey based data, hence their analysis require new methods. To this end, promising advances in deep learning over the past five years have achieved close-to-human performance on visual tasks such as recognizing objects and classifying them (e.g., telling a chair apart from a bird). In this project, I aim to apply advanced machine learning techniques to answer important questions in environmental health. Specifically, the focus will be on analysing images captured by satellites and street view cameras, and data generated from mobile phones, and investigating what they can reveal about health outcomes and its environmental and social determinants.
Technical Summary
Recent advances in deep learning methods have achieved unprecedented improvements in accuracy on multiple visual tasks such as object recognition and classification. While previously unused large-scale data is gaining importance in public health research, recent technical developments mostly focused on advanced spatial machine learning and statistics methods while integration of deep learning techniques have been largely unexplored. The overall goal of my fellowship research is to leverage these advanced methods to answer important questions in environment and health research. Specifically, convolutional neural networks will be trained using satellite and street view imagery to extract outcomes of health and its environmental/social determinants. Transfer learning and class specific saliency maps will be used for post-hoc model visualisation. A combined analysis of models trained for multiple outcomes will be conducted to study overlaps and deviations. Secondly, joint use of mobile phone and image data as predictors will be explored for improving model performance. New methods of data integration will be developed to enable the use of multiple big data sources as predictors in a unified modelling framework. Finally, transferability of deep learning models trained on data from one city to other geographies will be evaluated; adaptation techniques to facilitate transferability will be explored.
Publications
Bennett JE
(2023)
Changes in life expectancy and house prices in London from 2002 to 2019: hyper-resolution spatiotemporal analysis of death registration and real estate data.
in The Lancet regional health. Europe
Danesh Yazdi M
(2020)
Predicting Fine Particulate Matter (PM2.5) in the Greater London Area: An Ensemble Approach using Machine Learning Methods
in Remote Sensing
Jimenez MP
(2022)
Street-view greenspace exposure and objective sleep characteristics among children.
in Environmental research
Nathvani R
(2022)
Characterisation of urban environment and activity across space and time using street images and deep learning in Accra.
in Scientific reports
Sorek-Hamer M
(2022)
A Deep Learning Approach for Meter-Scale Air Quality Estimation in Urban Environments Using Very High-Spatial-Resolution Satellite Imagery.
in Atmosphere
Suel E
(2021)
Multimodal deep learning from satellite and street-level imagery for measuring income, overcrowding, and environmental deprivation in urban areas.
in Remote sensing of environment
Suel E
(2022)
What You See Is What You Breathe? Estimating Air Pollution Spatial Variation Using Street-Level Imagery
in Remote Sensing
Suel E
(2019)
Measuring social, environmental and health inequalities using deep learning and street imagery.
in Scientific reports
Description | Built Environment Assessment through Computer visiON (BEACON): Applying Deep Learning to Street-Level and Satellite Images to Estimate Built Environment Effects on Cardiovascular Health |
Organisation | Harvard University |
Department | Harvard T.H. Chan School of Public Health |
Country | United States |
Sector | Academic/University |
PI Contribution | Creating built environment exposure measures by leveraging deep learning algorithms applied to nationwide street imagery (2007-2020) to create fine-scale, time-varying built environment metrics of the natural environment (e.g., trees), physical environment (e.g., sidewalks), perceptions (e.g., safety), and urban form (e.g., compact high-rise). |
Collaborator Contribution | Funding secured for the project. Use a mix of innovative analytical approaches to determine the effect of the built environment on CVD-related health behaviors and CVD incidence across different time horizons. |
Impact | Not yet. |
Start Year | 2020 |
Description | CRESSH Built Environment Deep Learning Algorithms for Massachusetts (CRESSH-BEDLAM) Study |
Organisation | Harvard University |
Department | Harvard Medical School |
Country | United States |
Sector | Academic/University |
PI Contribution | Applying deep learning methods to generate novel measures built environment based on street level images throughout Massachusetts. |
Collaborator Contribution | Provision of funding for the work, selection of metrics most relevant for health studies, making the metrics available to be used by all CRESSH (Center for Research on Environmental & Social Stressors in Housing Across the Life Course) to enable their use in research by the community. |
Impact | No outputs yet - planned outputs for this year include datasets and publications. |
Start Year | 2019 |
Description | DeepInAfrica: Deep statistical learning based image analysis for measurement of socioeconomic development in Sub-Saharan Africa |
Organisation | ETH Zurich |
Country | Switzerland |
Sector | Academic/University |
PI Contribution | Co-Investigator for project funding from Swiss Data Science Center, Switzerland. Expertise in deep learning methods in applied settings for using street and satellite images. |
Collaborator Contribution | Application to Sub-saharan cities on identification of poor housing and slums. |
Impact | Project funding secured for hiring a postdoc (with collaborators) and two data scientists allocated to work on the project from the Swiss Data Science Center |
Start Year | 2021 |
Description | Environmental influences on Child Health Outcomes (ECHO) Program - Developing Google Street View-based Metrics of Nature and Their Influence on Health. |
Organisation | Harvard University |
Department | Harvard Medical School |
Country | United States |
Sector | Academic/University |
PI Contribution | Using deep learning algorithms and pre-trained networks for computing built environment metrics for cities for Boston and New York to be used in ECHO cohort studies. |
Collaborator Contribution | Funding secured for the project. Estimating associations of image derived natural environment metrics with health using three ECHO cohorts (Project Viva, ACCESS, and PRISM). Health outcomes include prenatal depression, physical activity and sleep, adiposity, asthma, insulin resistance, and positive health. Disseminate the software toolkit to be used by other researchers and ECHO cohorts. |
Impact | No outputs yet - outputs to be reported next year. |
Start Year | 2019 |
Description | Environmental influences on Child Health Outcomes (ECHO) Program - Developing Google Street View-based Metrics of Nature and Their Influence on Health. |
Organisation | Oregon State University |
Country | United States |
Sector | Academic/University |
PI Contribution | Using deep learning algorithms and pre-trained networks for computing built environment metrics for cities for Boston and New York to be used in ECHO cohort studies. |
Collaborator Contribution | Funding secured for the project. Estimating associations of image derived natural environment metrics with health using three ECHO cohorts (Project Viva, ACCESS, and PRISM). Health outcomes include prenatal depression, physical activity and sleep, adiposity, asthma, insulin resistance, and positive health. Disseminate the software toolkit to be used by other researchers and ECHO cohorts. |
Impact | No outputs yet - outputs to be reported next year. |
Start Year | 2019 |
Description | Explore 2021: Discover The Power Of Global Connection |
Form Of Engagement Activity | A talk or presentation |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Professional Practitioners |
Results and Impact | Online conference organised by users of satellite images that are produced by a private company - users include governments, researchers, NGOs, other practitioners. |
Year(s) Of Engagement Activity | 2021 |
Description | Invited panelist (Salzburg Global Seminar) |
Form Of Engagement Activity | Participation in an activity, workshop or similar |
Part Of Official Scheme? | No |
Geographic Reach | International |
Primary Audience | Third sector organisations |
Results and Impact | Meeting with people working in policy with regards to protection of natural environments in cities. I was invited to attend a panel discussion on use of data and technologies in urban environments, sparking questions and discussions both during the session and afterwords with policy makers, third party organisations, and start-up businesses. |
Year(s) Of Engagement Activity | 2019 |
URL | https://www.salzburgglobal.org/multi-year-series/parks/pageId/session-620.html |