RAPID IDENTIFICATION OF STREPTOCOCCUS AND ENTEROCOCCUS SPECIES USING DIFFUSE REFLECTANCE-ABSORBENCY FOURIER-TRANSFORM INFRARED-SPECTROSCOPYAND ARTIFICIAL NEURAL NETWORKS

Diffuse reflectance-absorbance Fourier transform infrared spectroscopy (FT-IR) was used to analyse 19 hospital isolates which had been ident ified by conventional means to one of Enterococcus faecalis, E. faeciu m, Streptococcus bovis, S. mitis, S. pneumoniae, or S. pyogenes. Princ ipal components analysis of the FT-IR spectra showed that this 'unsupe rvised' learning method failed to form six separable clusters (one for each species) and thus could not be used to identify these bacteria b ased on their FT-IR spectra. By contrast, artificial neural networks ( ANNs) could be trained by 'supervised' learning (using the back-propag ation algorithm) with the principal components scores of derivatised s pectra to recognise the strains from their FT-IR spectra. These result s demonstrate that the combination of FT-IR and ANNs provides a rapid, novel and accurate bacterial identification technique.