Spectroscopic Imaging for Detection and Discrimination of Different E. coli Strains

Food contaminations with E. coli bacteria are a major concern for public health. Current techniques for detection are based on sample extractions. time-consuming sample preparations, and labor intensive analyses. Because some strains can be toxic at a level of tens of bacteria and sonic are not harmful at all, a method of colony localization and strain classification must be developed. In this study we present first results that are based on Fourier transform infrared (FT-IR) spectroscopy and FT-IR imaging. Due to the chemical similarity of different E. coli strains, the acquired spectra show a strong resemblance. It is demonstrated here that based on a correlation analysis samples of the same strain are classified as such and that different strains can be discriminated. The next step is to move from single-spot analyses towards spectroscopic imaging-a technique that facilitates detection of localized bacteria colonies. However, the sheer amount of data acquired in short periods of time prevents many chemical imaging techniques from being feasible for online sensing or for screening extended areas. To improve the time resolution, a data compression approach based on three-dimensional wavelet compression has been applied. It is shown that even with slight compression computation times can be cut down by over an order of magnitude while preserving enough information for localization and classification.