Laser-induced breakdown spectroscopy coupled with machine learning for rapid quantification of Escherichia coli concentration

• Published in: Talanta (Oxford) Vol. 296; p. 128522
• Main Authors: Wang, Jingjing, Liang, Jiahui, Chen, Fei, Yu, Runheng, Tian, Zhihui, Zhao, Yang, Ma, Weiguang, Dong, Lei, Li, Jiaxuan, Yin, Wangbao, Xiao, Liantuan, Jia, Suotang, Zhang, Lei
• Format: Journal Article
• Language: English
• Published: Netherlands Elsevier B.V 01.01.2026
• Subjects: Bacterial concentration; Escherichia coli; Escherichia coli - isolation & purification; Laser-induced breakdown spectroscopy; Lasers; Machine Learning; Spectrum Analysis - methods; Support Vector Machine; Support vector regression; Support vector regression; Escherichia coli; Laser-induced breakdown spectroscopy; Bacterial concentration; Machine learning
• ISSN: 0039-9140; 1873-3573; 1873-3573
• DOI: 10.1016/j.talanta.2025.128522

The rapid and accurate quantification of bacterial concentrations is essential for food safety monitoring, environmental surveillance, and clinical diagnostics. Traditional methods are often limited by lengthy procedures, complex operations, or high costs. This study developed a novel approach combing laser-induced breakdown spectroscopy (LIBS) with machine learning for rapid bacterial concentration analysis. Using Escherichia coli (E. coli) as a model organism, we systematically optimized key LIBS parameters including delay time, substrate material, and laser repetition rate to achieve optimal spectral quality. Three machine learning algorithms - support vector regression (SVR), gradient boosting regression (GBR), and kernel ridge regression (KRR) – were comparatively evaluated. The SVR model demonstrated superior performance with a coefficient of determination (R2) of 0.99, along with root mean square error (RMSE) of 7.3 × 105 cells/mL and mean absolute error (MAE) of 4.2 × 105 cells/mL, respectively. Method validation showed recovery rates ranging from 100.03 % to 100.83 %, with relative standard deviations (RSD) less than 2 %. The t-test confirmed no significant difference between the spiked concentrations and the detected concentrations (p > 0.05), indicating that the method possesses excellent accuracy and precision. This multi-feature integration approach effectively addressed the nonlinear correlation between spectral line intensity and bacterial concentration in LIBS quantification. The method offers significant advantages including minimal sample preparation and rapid analysis speed. These findings establish a reliable and efficient technique for microbial quantification with promising applications in food production facilities, healthcare settings, and ecological studies. [Display omitted] •A LIBS-based method combined with machine learning was proposed for rapid bacterial concentration quantification.•Accurate quantification was achieved via optimized parameters and algorithm comparison.•Multi-feature fusion effectively addressed the non-linear relationship between spectral intensity and concentration.