Detection of airborne virus particles using hyperspectral dark field microscopy and machine learning

Abstract The development of technologies that rapidly screen airborne respiratory viruses providing near real-time data is essential for swift assessment and response to potential outbreaks, ultimately protecting public health. However, many current technologies require laboratory settings, specialised equipment, and trained personnel, which limits their applicability for widespread and real-time monitoring. Implementing these systems can be cost-prohibitive and logistically complex, especially in low-resource settings, highlighting the need for easily deployable technologies. This paper presents a novel proof-of-concept technology for near-real-time screening of specific viral targets, employing a machine-learning-enhanced optical system to analyse hyperspectral scattering data from airborne virus particles. This approach enables non-destructive and label-free identification of specific morphological particle classes by using enhanced dark-field microscopy to capture unique spectral signatures of the captured targets, which are then classified by a machine-learning algorithm. We demonstrate the laboratory-validated potential of this technology for rapid screening of airborne viral proxies, including inactivated SARS-CoV-2 as a target, offering a promising preliminary screening approach for indoor anomaly detection and potential early warning indicators.