Decoding Physical and Cognitive Impacts of PM Concentrations at Ultra-fine Scales

Abstract

The human body is an incredible and complex sensing system. Environmental factors trigger a wide range of automatic neurophysiological responses. Biometric sensors can capture these responses in real time, providing clues to the underlying biophysical mechanisms. Here we show biometric variables can be used to accurately estimate ultra-local particulate matter concentrations in the ambient environment with high fidelity (r$^2$ = 0.91) and that smaller particles are better estimated than larger ones. Inferring environmental conditions solely from biometric measurements allows us to disentangle key interactions between the environment and the body. A deeper understanding of these interactions can have countless important applications in public health, preventative healthcare, city planning, human performance, and much more. By tapping into our body's `built-in' sensing abilities, we can gain insights to how our environment influences our physical health and cognitive performance.

Gauging Size Resolved Ambient Particulate Matter Concentration Solely Using Biometric Observations: A Machine Learning and Causal Approach

“…Multiple trials may improve the predictive power of the model, similar to the previous single-subject experiment. 14 Through the usage of SHAP values, it is also revealed that skin temperature and heart rate hold the most importance for this empirical ML model. Research has shown that there is a relationship between skin temperature and heart rate in children, 26 but the short-term effects of PM concentration on physiological responses such as skin temperature and heart rate are still being explored.…”

“…Empirical ML models that predict PM concentration on multiple participants extend the results that predicted PM concentration from the biometric signature of a single participant. 14 The first model, Figure 6, included all biometric characteristics collected and derived, represented by 324 columns, ranging from EEG frequency spectral densities to pupillometrics to skin temperature and heart rate. This is an improvement over the previous results from a single subject, which had a goodness of fit of r 2 = 0.91 between the predicted and actual PM concentrations for the test data set.…”

“…This study builds on the earlier study of 14 by using a much larger data set amassed from a diverse cohort of participants, substantially increasing the number of data points available for the development of machine learning models, and expands the analysis to include a causal perspective and derive a directed graph based on the collected data that describes the relationship between variables. This causal perspective highlights the importance of physiological responses as indicators of exposure to particulate matter, providing a basis for the use of biometric data in environmental health surveillance and real-time pollution assessment.…”

“…The data collection methodology is inherited from the previous iteration of this study. 14 Although the fundamental methodologies for data collection, described in Sections 2.1 and 2.2, have remained consistent, the main differences and improvements of the study presented are outlined below.…”

“…12 Among patients with chronic obstructive pulmonary disease (COPD), short-term exposure to PM2.5 can result in decreased blood oxygen saturation (SpO2) in the order of minutes. 13 This manuscript describes an extension of a previous study, 14 and contains biometric data from four subjects collected in four cases in October 2021, January, and February 2022. We observe that with additional data from increased subjects, machine learning (ML) models can better predict smaller PM concentrations from a handful of biometric features (skin temperature, heart rate, and respiratory rate).…”