Rapid and In-situ Detection of Biological Agents via Near-Infrared Spectroscopy and Cloud-Powered Neural Networks
The contagious nature of certain biological agents and the difficulty in treating infections render them a significant threat to public health and safety. In situations involving chemical, biological, radiological, and nuclear (CBRN) agents, effective detection is of paramount importance to prevent the undetected spread of these agents and enable swift and targeted responses tailored to the specific threat. Although portable detection tools are effective for chemical and nuclear detection, current biological detection methods face several challenges, including limited mobility, extended processing times, and varying accuracy. In the context of biological threats, where agents such as anthrax can be rapidly dispersed due to environmental factors and human activities, rapid detection is of paramount importance. It is imperative to develop field-applicable detection devices that are highly selective, capable of differentiating between biological and non-biological agents, as well as benign and harmful microorganisms. This study examines the potential of near-infrared (NIR) spectroscopy in conjunction with machine learning as a rapid in-situ biological detection method. The objective is to distinguish between biological agents and common white powders that are used as confounding agents in suspect letters. The non-pathogenic surrogates employed are safe and representative of typical biological warfare agents. The near-infrared (NIR) spectra of lyophilized bacterial and fungal surrogates, along with common white powders, were subjected to analysis and processing through the application of principal component analysis (PCA) and hierarchical clustering analysis (HCA). This resulted in the successful classification of the samples into distinct groups. The classification model demonstrated high accuracy in its prediction, thereby emphasizing the potential of the method for field detection of solid biological agents. These promising results suggest that NIR spectroscopy combined with machine learning could be further investigated as a rapid in-situ tool for biological detection in CBRN contexts.