Label-free detection of pathogenic microorganism using Ag NPs@PDMS sponge SERS substrate and machine learning

Abstract

This work aims to face the challenge of label-free and quantitative detection for pathogenic microorganisms concerning on food safety and medical diagnosis. Ag NPs@PDMS sponge SERS substrates were developed through the "mold transfer-surface embedding" method for pathogenic microbial detection. The pore structure provided large specific surface area and critical contact sites, which endowed the detection limit of only 1 CFU/mL for E.coli. Importantly, six different microorganisms could be identified using this SERS substrate. The classification models combining three types of data dimension reductions and five types of machine learning algorithms were constructed to achieve the automatic classification of microorganisms. Among these algorithms, the PCA-SVC model held the highest accuracy of 92.17%. Different from labeled SERS, label-free SERS technique is conducive to the discovery of unknown microorganisms in water-environment. Therefore, this SERS substrate is expected for microbial system assessment and microbial safety monitoring, combined with handheld Raman spectrometer and artificial intelligence.