Abstract
Detection of pathogens is a major concern in many fields like medicine, pharmaceuticals, or agri-food. Most conventional detection methods require skilled staff and specific laboratory equipment for sample collection and analysis or are specific to a given pathogen. Thus, they cannot be easily integrated into a portable device. In addition, the time-to-response, including the sample collection, possible transport to the measurement equipment, and analysis, is often quite long, making real-time screening of a large number of samples impossible. This paper presents a new approach that better fulfills industry needs in terms of integrated real-time wide screening of a large number of samples. It combines optical imaging, object detection and tracking, and machine-learning-based classification. Three of the most common bacteria are selected for this study. For all of them, living bacteria are distinguished from inert and inorganic objects (1 μm latex beads) based on their trajectory, with a high degree of confidence. Discrimination between living and dead bacteria of the same species is also achieved. Finally, the method successfully detects abnormal concentrations of a given bacterium compared to a standard baseline solution. Although there is still room for improvement, these results provide a proof of concept for this technology, which has strong application potential in infection spread prevention.