Ultra-fast identification of lactic acid bacteria colonies based on droplet microcavity label-free SERS

In this study, we addressed the challenge of excessive fluorescence background in bacterial colony Raman detection and aimed to achieve rapid identification of colonies. To overcome this issue, we employed a combination of droplet microcavity and label-free Surface Enhanced Raman Spectroscopy (SERS) technologies for spectroscopic analysis of five species of lactic acid bacteria (LAB) colonies during fermentation. This approach, coupled with Supported Vector Machine (SVM) and K-Nearest Neighbors (KNN) machine learning algorithms, facilitated the identification and analysis of spectral data. Comparing the results with conventional bacterial colony Raman spectra, the SERS spectra exhibited clear peaks, a higher and more stable signal-to-noise ratio, and noticeable spectral differences between various colonies, overcoming the limitations of insufficient fluorescence background. Moreover, the detection speed was notably enhanced, each SERS spectrum requires only 0.5 s, and the acquisition of the 100 spectral data points necessary for one bacterial colony is accomplished in less than 1 min. The SVM algorithm demonstrated a bacterial colony identification rate exceeding 95%, while the KNN algorithm achieved a rate surpassing 90%. These findings highlight the practical importance of using droplet microcavity combined with label-free SERS technology for quick and robust identification of the bacterial colonies.