Abstract
Exposure assessment studies are the primary means for understanding links between exposure to chemical and physical agents and adverse health effects. Recently, researchers have proposed using wearable monitors during exposure assessment studies to obtain higher fidelity readings of exposures actually experienced by subjects. However, limited research has been conducted to link a wearer’s actions to periods of exposure, a necessary step for estimating inhaled dosage. To aid researchers in these settings, we developed a machine learning model for identifying periods of bicycling activity using passively collected data from the RTI MicroPEM wearable exposure monitor, a lightweight device capable of continuously sampling both air pollution levels and accelerometry parameters. Our best performing model identifies biking activity with a mean leave-one-session-out (LOSO) cross-validation F1 score of 0.832 (unweighted) and 0.979 (weighted). Accelerometer derived features contributed greatly to the model performance, as well as temporal smoothing of the predicted activities. Additionally, we found competitive activity recognition can occur with even relatively low sampling rates, suggesting suitability for exposure assessment studies where continuous data collection for long periods (without recharge) are needed to capture realistic daily routines and exposures.
Highlights
Mitigating human exposure to air pollution is a top public health objective
1 − pi where pi is the probability that outcome is positive, xi is a vector of covariate values for observation i and w is a vector of weights shared across all observations, learned from the data
While we hypothesized that non-accelerometer readings could help distinguish periods of biking, their contribution paled in comparison to the impact of the accelerometer readings
Summary
Mitigating human exposure to air pollution is a top public health objective. According to the most recent global burden of disease/comparative risk assessment, exposure to air pollution causes over 6.8 million premature deaths per year, with roughly equal contributions from ambient and household air pollution [1]. While many countries acknowledge this danger and have enacted air pollution prevention laws, developing specific rules and regulations to implement these laws can be challenging when the relationships between pollutant exposure and health effects are underspecified. Better information on exposure response relationships (how clean is clean enough?) can help regulators more effectively balance the often competing objectives of economic growth and public health. Exposure assessment studies are the primary means for understanding links between exposure to chemical and physical agents and adverse health effects. Exposure assessment studies have traditionally used imperfect proxies based on central site data to assess these relationships, even though ambient concentrations are poor predictors of Sensors 2019, 19, 4613; doi:10.3390/s19214613 www.mdpi.com/journal/sensors
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
