| Listeriosis is a serious food-borne illness. Current methodology requires five to seven days to confirm the presence of Listeria monocytogenes , the causative agent. Pyrolysis/gas chromatography offers a rapid analytical technique that generates a prodigious amount of data. In this research the patterns generated by the pyrolysis of various Listeria species are analyzed without regard to the biochemical compounds causing these patterns. The eventual goal is to speciate these organisms by computational analyses of these pyrolysis patterns.; Initially, the experimental parameters are presented The columns examined are DB-5, DB-17, and DB-Wax. The raw chromatographic data consists of a data file with 15,000 data points representing almost fifty minutes of run time. A processing scheme is presented which eliminates those portions of the chromatogram, which are information poor and standardizes the scaling. This technique produces data files of 10,000 data points. Although the above mentioned processing reduces the data files, 10,000 is still rather large for analytical purposes. Another technique to reduce the dimensionality of these data sets is wavelet transform. Wavelet transform is a mathematical tool, which decomposes a signal into an approximation signal and a detail signal each half as long as the original signal. If the approximation signal is used in place of the original the length is cut in half The wavelet transform can then be performed on the approximation signal in an iterative manner. In the current study wavelet transform is used to reduce the size of the data files from 10,000 to just over 600.; Finally, artificial neural networks are presented A general background and method of use is briefly outlined. Probabilistic neural networks (PNNs) are presented in more detail, as these are the specific neural networks used for pattern recognition The two data sets of wavelet coefficients (DB-5 and DB-17) are used as training sets for constructing and testing various PNN models. The results of a manual cross validation method of testing the various PNNs demonstrate the utility of this method for speciating Listeria . The method provides a highly flexible tool for identifying these organisms, which can be adjusted to provide one of several different levels of accuracy depending upon the ultimate goal of use. |
