Study On Direct Identification Of Bacteria By Laser-induced Breakdown Spectroscopy

Bacteria are everywhere in the natural environment.Although most of them are harmless or even beneficial,there are still some hazardous bacteria that will harm human health,such as the well-known Staphylococcus aureus,Xanthomonas Campestris pv.oryzae,Yersinia pestis and so on.Therefore,it is particularly important to identify bacteria quickly and accurately.Existing traditional biological identification methods such as morphological identification,Polymerase Chain Reaction(PCR),16 S r DNA and so on,although the identification rate is high,but time-consuming and laborious and the process is complicated.Laser-induced breakdown spectroscopy(LIBS)is one of the potential technologies for rapid identification of bacteria,due to its benefits of real-time rapid measurement,nearly little sample preparation and multi-element measurement.In this paper,six weakly active bacteria,including Escherichia coli,Enterococcus faecalis,Bacillus megaterium,Bacillus thuringiensis,Pseudomonas aeruginosa and Bacillus subtilis,were taken as analysis samples.We adopted a simple and convenient sample pretreatment method to treat bacteria samples,which was simpler and faster than the traditional culture,centrifugation,lyophilization and other treatment methods.The treated bacterial samples were evenly coated on five high-purity substrates of high purity graphite,silicon,aluminum,zinc and stannum for drying.In terms of spectral data processing,the experimental substrate was optimized and the feasibility of LIBS identification of these bacteria was verified.Normalization and Principle Component Analysis(PCA)were combined for preprocessing the spectrum collected by the experiment,and Support Vector Machine(SVM)based on one-against-one and linear kernel function was established to model and predict the spectrum of experimental acquisition.Finally,accuracy,recall and identification rate were used to evaluate the identification performance of the model.This paper firstly introduced the research status of traditional bacterial identification and LIBS identification bacteria,the principle and characteristics of LIBS,and then described the preparation method of bacterial samples and the LIBS experimental equipment,and then introduced the basic principle of PCA,SVM and other algorithms used in spectral processing.Finally,the collected spectrum was preprocessed,and SVM was used for spectrum modeling and prediction analysis,and the prediction results were evaluated by accuracy,recall and identification rate.The innovations in this article was:By laser-induced breakdown spectroscopy,a series of procedures for the precise and direct identification of bacteria had been created.The high-energy laser pulse were directly ablatesd the surface of bacterial samples to generate plasma.Normalization and principal component analysis(PCA)were used to preprocess the spectrum,and a multi-class identification method based on the one-agianst-one linear kernel function of support vector machine(SVM)combined with the characteristic spectral lines of Ca,Mg,Fe were proposed for modeling and prediction.The identification rate of six bacteria,such as Escherichia coli,Enterococcus faecalis,had been determined to be 89.17%.The thesis was divided into the following parts based on the above content:1.An introduction of the classification,application and current research status of bacterial identification in biology was provided,and the detection principle of LIBS and the research status of LIBS identification bacteria were expounded.2.The experimental setup for LIBS bacterial identification,the process of bacterial sample preparation,and working environment for experimental operations were elaborated,and the experimental substrate optimization approach was presented.3.The basic principles and application processes of various spectral processing algorithms such as normalization,PCA and SVM were described,and the evaluation methods of the identification effect of bacterial species by using precision,recall and identification rate were imported.4.The effects of pretreatment of bacterial spectrum using various algorithms were compared and optimized,experiment outcomes of complete spectrum and partial spectrum were evaluated.SVM was further used to train and model the preprocessed sample data,which was utilized to predict the data of prediction set.Ultimately,the identification performance of bacterial species was assessed with precision,recall and identification rate.