Previously, we successfully realized the identification of a single species of bacteria based on the multi-wavelength transmission spectrum of bacteria. The current research is focused on realizing the spectral analysis of mixed bacteria. Principal component analysis-Monte Carlo (PCA-MC) model was developed for the implementation of spectral separation of mixed bacteria by obtaining the ratio of components. And, the separated spectrum was regarded as the model input of the neural network concentration inversion model to obtain the concentration of each bacteria in the mix. Mean relative errors in component analysis of mixing S.aureus with K.pneumoniae, mixing S.aureus with S.typhimurium twice, mixing K.pneumoniae with S.typhimurium are 3%, 2%, 3.9% and 6.1%, respectively. The coefficient of determination (R2) of validation set and test set are 0.9947 and 0.9954 in concentration inversion model. The results show that this method can quickly and accurately determine the component ratio and concentration information in the mixed bacteria. A new method was proposed to separate the spectrum of mixed bacteria effectively and measure its concentration quickly, which makes a big step forward in the detection and online monitoring of waterborne microbial contamination based on multi-wavelength transmission spectroscopy.
Keywords: Backpropagation neural networks; Bacteria; Monte carlo; Multi-wavelength transmission spectrum; Principal component analysis; Spectral separation.
Copyright © 2022 Elsevier B.V. All rights reserved.