Discrimination of hazardous bacteria with combination laser-induced breakdown spectroscopy and statistical methods

Real-time biohazard detectors must be developed to facilitate the rapid implementation of appropriate protective measures against foodborne pathogens. Laser-induced breakdown spectroscopy (LIBS) is a promising technique for the real-time detection of hazardous bacteria (HB) in the field. However, distinguishing among various HBs that exhibit similar C, N, O, H, or trace metal atomic emissions complicates HB detection by LIBS. This paper proposes the use of LIBS and chemometric tools to discriminate Staphylococcus aureus, Bacillus cereus, and Escherichia coli on slide substrates. Principal component analysis (PCA) and the genetic algorithm (GA) were used to select features and reduce the size of spectral data. Several models based on the artificial neural network (ANN) and the support vector machine (SVM) were built using the feature lines as input data. The proposed PCA-GA-ANN and PCA-GA-SVM discrimination approaches exhibited correct classification rates of 97.5% and 100%, respectively.