Development of a hierarchical classification system with artificial neuralnetworks and FT-IR spectra for the identification of bacteria

The practical value of elaborated vibrational spectroscopic techniques in m edical and microbiological biodiagnostics depends strongly on the reliabili ty, the speed, the ease of use, and the evaluation procedures of the acquir ed data. Ln the present study, artificial neural networks (ANNs) were used to establish a hierarchical classification system for microbial Fourier tra nsform infrared (FTIR) spectra suitable for identification purposes in a ro utine microbiological laboratory: A radial basis function network (RBF) pro ved to be superior for a top-level classification of the FT-IR spectra at t he genus level. Species within these genera were sequentially further class ified by using multilayer perceptrons (MLPs), which achieved a larger diffe rentiation depth than RBF networks. The MLPs were trained with several lear ning algorithms. Best performance was achieved with the cascade correlation (CC) approach to determine the network topology combined with resilient pr opagation (Rprop) as the training algorithm. The final hierarchically organ ized model was able to discriminate between four genera of microorganisms c omprising 42 different strains of Pseudomonacae, 33 strains of Bacillus, 46 strains of Staphylococcus, and 6 species and 24 strains of yeast genera Ca ndida. Altogether, 145 strains from international microbial strain collecti ons are comprised in 971 spectra. The species Candida albicans could be fur ther classified with respect to susceptibility against the antibiotic drug fluconazole, which is of therapeutic relevance. Key factors for the classif ication results of the bacterial FT-IR spectra were the data pretreatment, the number of wavelengths selected by a feature extraction algorithm, the t ype of network, and the learning function used for the ANN training.