Novel Applications for Bioaerosol Classification: An in-depth Analysis using a Wideband Integrated Bioaerosol Sensor

Abstract

Bioaerosols, which consist of airborne particles containing living organisms or their derivative components, have significant implications across various environmental and societal components. They impact human health, contribute to agricultural productivity, and influence atmospheric processes. Traditional methodologies for studying bioaerosols have predominantly been anchored in laboratory-based techniques, including microbial culturing and microscopy, among others. The advent of real-time Ultra Violet Light-Induced Fluorescence (UVLIF) instrumentation, exemplified by devices such as the Wideband Integrated Bioaerosol Sensor (WIBS), has transformed the field. These instruments facilitate in-situ measurements, capturing an array of particle attributes like size, shape, and fluorescence profiles. Conventionally, the data procured from UVLIF instruments have been subject to analytical scrutiny using machine learning techniques such as K-Means and Hierarchal Cluster Analysis (HCA), predominantly in the context of large-scale ambient environmental studies. However, recent years have seen a renewed interest in leveraging UVLIF technology for indoor air quality assessments, possibly in response to the global COVID-19 pandemic. The impetus for this interest is twofold: the urgent need to understand the dynamics of aerosolized pathogens in various indoor settings, and the requirement for real-time monitoring to inform and improve public health responses. UVLIF technology offers a non-invasive, rapid assessment of particulate matter in the air by detecting the characteristic fluorescence emitted by biological particles when exposed to UV light. This capability is particularly pertinent given the evidence suggesting that SARS-CoV-2 can be transmitted via aerosols, which may linger and accumulate, especially in poorly ventilated areas. Consequently, real-time monitoring of these aerosols becomes essential for evaluating the risk of airborne transmission and the effectiveness of interventions such as air filtration, ventilation, and social distancing measures. The renewed interest in UVLIF for indoor air quality is also reflective of a broader trend towards more proactive and preventive measures in public health, particularly in mitigating the spread of airborne infectious diseases. Thus, applicability of these instruments has expanded significantly, opening new avenues for research in a burgeoning scientific domain. Despite the versatility of UVLIF instruments and the advantages they offer, traditional clustering algorithms like K-Means and HCA present distinct limitations, particularly when applied to real-time or ‘online’ data analyses. For example, K-Means operates under the assumption of spherical cluster geometry and lacks effective mechanisms for outlier management. Conversely, HCA, although offering more intricate analyses, comes with drawbacks of computational expense and sensitivity to noise. In this thesis, we venture into exploring the innovative application of the WIBS instrument within a clinical setting to investigate the fluorescent properties of aerosols produced by human participants during a variety of speech and language therapy activities. Additionally, using data obtained from a large environmental campaign, we shall evaluate both traditional and emergent unsupervised machine learning techniques to establish their efficacy and suitability in a rapidly advancing field. In the third paper, we will extend the scope of our study to incorporate a subsequent environmental campaign the following year, providing a year-on-year comparison of bioaerosol concentrations and morphological attributes – an area surprisingly underdeveloped in existing literature. Lastly, the final chapter of the thesis shall survey the challenges and opportunities involved in curating a robust catalogue of particles for utilisation in supervised learning paradigms.

Date of Award	1 Aug 2024
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Martin Gallagher (Supervisor) & David Topping (Supervisor)