Leveraging AI/ML for Environmental Health Risk Assessment in Homes
openNIEHS - National Institute of Environmental Health Sciences
PROJECT SUMMARY / ABSTRACT
Millions of people in the United States experience respiratory health conditions such as asthma, COPD, and
allergies that make them more susceptible to the acute health impacts of indoor air pollution, including
coughing, difficulty breathing, wheezing, dizziness, and irritation. Interventions are needed to reduce acute
exposures and health risks in homes to (1) identify sources of pollution and (2) guide residents in effective
exposure mitigation strategies. Sources include emissions from smoking, fires, cooking, cleaning, dust,
household products, and ambient pollution. Reducing indoor air exposure will also reduce long-term risks such
as cancer and cardiovascular diseases. To meet these needs, we will leverage the latest sensor technology,
voice and cloud platforms, and Artificial Intelligence / Machine Learning (AI/ML) technologies to develop a Just-
in-Time Adaptive Intervention for households, called CLYR-VUE, that learns precise exposure contexts and
delivers context-aware real-time guidance in the moments that hazards occur. The intervention will help users
verify and maximize intervention effectiveness with adaptive refinements over time. We will first develop the
hardware and software infrastructure needed to implement the CLYR-VUE intervention. We will use industry-
standard sensors for environmental quantities that can be used to indicate different sources (particle counts by
size, VOCs, CO, CO2, NO2, HCHO, temperature, humidity, light, and noise). Next, we will evaluate and test its
performance in the laboratory and conduct an intensive field pilot test in 4 real homes with adults and children
to evaluate its feasibility in terms of usability, acceptability, technical performance, and potential health impact.
Our specific aims with tasks are: AIM #1. Develop software and hardware (Engineer and build 25 units →
Add Voice Interface → Web scraping of public air quality data → AI/ML computation → Messaging engine →
Web app frontend). AIM #2. Perform 73 laboratory tests (Measure source emissions in a chamber →
Evaluate and calibrate sensor readings → Train and evaluate ML models for source classification). AIM #3.
Perform pilot field test in 4 homes (Develop procedures → Deploy 2-5 devices per home → Evaluate CLYR-
VUE feasibility). We will elicit feedback from participants on each component of their experience, including
setup, instruction, voice interaction, web app use, accuracy of source identification, appropriate messaging,
guidance, impact on behavior and health, cost, and privacy. We will also observe objective changes in
exposure events and mean pollutant levels as a measure of potential impact. This study will feed into a Phase
II impact study and commercialization.
Up to $307K
health research