The identification of Listeria Monocytogenes based on the electronic nose

Our previous work on electronic nose (E-nose) could effectively detect several major foodborne pathogens, which was based on principle component analysis (PCA) and cluster analysis (CA) method. But these methods could not identify two strains of Listeria Monocytogenes, which had the serotypes of 4b and 4c respectively. To resolve this problem, we proposed artificial neural network method which could differentiate these two strains. The specifically constructed neurons could get feature data from the E-nose output text files. The first layer could detect the volatile metabolites of 4 species of Listeria spp. and 9 strains of L. monocytogenes. This was because the volatile metabolites in the culture medium of 4 species of Listeria spp. could be well distinguished by PCA. The second layer could use the selected differential feature data to identify 4b and 4c serotypes of L. monocytogenes. Experiment showed that the improved E-nose method had good stability and repeatability. This study indicates that the odor fingerprint based on detecting microbial volatile metabolites can be enhanced with new feature extracted by artificial neural network method and can be used in pathogen identification in future.