Research On Identification And Classification Of Microbial Single Cells Based On Multi-point Scanning Raman Spectroscopy

Raman spectroscopy has been widely used for microbial analysis due to its exceptional qualities as a rapid,simple,non-invasive,reproducible,and real-time monitoring tool.The Raman spectrum of a cell is a superposition of the spectral information of all biochemical components in the laser focus.Microorganisms vary in size and the laser spot size is usually constant.In cases where the size of individual microbial cells is larger than the laser spot size,Raman spectra measured from a single point within the cell cannot capture all biochemical information due to the spatial heterogeneity of microorganisms,which can affect the accuracy of microbial identification.In this paper,we have developed a multi-point scanning measurement method for Raman spectroscopy of microbial cells by introducing an image recognition algorithm that can adequately measure all biochemical component information within the microorganism.Multi-point scanning confocal Raman spectra had high recall rates(reaching 90%)and identification accuracy rates(~96%)compared to single-point Raman spectroscopy.The challenge is solved by the fact that the individual microbial cells being measured vary in size and the laser spot of the confocal Raman device cannot be easily changed.Combined with a wide range of statistical analysis tools,multi-point scanning confocal Raman spectroscopy can be used to more accurately classify species at different taxonomic levels,which is of great importance for species identification.The research within the thesis and the results obtained consist of three main parts.(1)Research spatial heterogeneity of microorganisms based on Raman spectroscopy.The data were first collected from individual cells of three microorganisms(Escherichia coli,Bacillus subtilis and Saccharomyces cerevisiae)using Raman mapping and the data were pre-processed to obtain Raman images of the three microorganisms in single cells.Analysis of Raman mapping images of three microorganisms and further analysis of the mapping data using Multivariate Curve Resolution-Alternating Least Squares(MCR-ALS)showed slight differences in the Raman spectra of individual cells at different locations showed slight differences,i.e.spatial heterogeneity.(2)To address the problem of spatial heterogeneity of microorganisms,a scanning confocal Raman spectroscopy technique based on image recognition algorithm is proposed,which enables Raman spectroscopy to measure multi-point biochemical composition information of microorganisms in a single exposure through the control of image recognition algorithm and high-precision motorized displacement stage,to solve the challenge that the measured individual microorganism cells are of different sizes and the laser spot of confocal Raman equipment is not easy to be changed so that Raman spectroscopy can become a real "fingerprint spectroscopy" of single cells.(3)To verify that multi-point scanning confocal Raman spectroscopy can fully obtain information on the internal cellular components and is beneficial to improve the accuracy of species identification,single-point Raman spectra and multi-point scanning Raman spectroscopy data of various microorganisms combined with machine learning algorithms(SVM,LDA,KNN,etc.)were analyzed and compared.The results show that multi-point scanning confocal Raman spectroscopy can comprehensively describe all biochemical information inside cells,and combined with extensive statistical analysis,can develop a more reliable database of microbial Raman spectra,which has great potential for microbial identification and classification applications.