Real-time detection and analysis of foodborne pathogens via machine learning based fiber-optic Raman sensor

Food safety is a critical concern worldwide, and in recent years, it has become a growing public concern in the United States due to mass production and multi-channel distribution of high-nutrient and fresh-cut foods that have increased the prevalence and diversity of foodborne pathogens (e.g., Escherichia coli, Salmonella, Listeria, etc.). Traditional methods for the detection of these pathogens rely heavily on low-efficiency techniques, which may expose the public to contaminated foods that have not been tested or identified. Therefore, rapid, simple, sensitive, and inexpensive detection methods are urgently needed for food safety investigations with higher efficiency and accuracy. One of the promising methods for the detection of foodborne pathogens is Raman spectroscopy, which is a molecular-level analytical tool with the advantages of high selectivity and sensitivity and simple operation at a relatively low cost. In this study, a novel fiber-optic-based portable Raman probe was developed and used for real-time detection of a panel of pathogen-specific molecular fingerprint volatile organic compounds (VOCs). Furthermore, machine learning (ML) algorithms were applied to assist in the extraction of molecular information from the raw Raman spectra, resulting in high accuracy prediction of complex VOC mixtures, even at high dilution folds (100x). The developed ML-assisted Raman probe has immense potential for rapid on-site detection of complex chemical mixtures in food safety and beyond. This innovative approach achieves high accuracy in comprehensively sorting mixed chemicals with varying concentrations and provides several advantages over previous studies, including speed, portability, non-contact operation, and precise classification of foodborne pathogen VOCs.