OPTICAL DETECTION OF FOOD SAFETY AND FOOD DEFENSE HAZARDS
Location: Quality and Safety Assessment Research Unit
Title: The effect of regions of interest and spectral pre-processing on the detection of non-O157 shiga-toxin producing escherichia coli serogroups on agar media by hyperspectral imaging
Submitted to: Near Infrared Spectroscopy Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: May 7, 2012
Publication Date: June 28, 2012
Citation: Windham, W.R., Yoon, S.C., Ladley, S.R., Heitschmidt, G.W., Lawrence, K.C., Park, B., Narang, N., William, C.C. 2012. The effect of regions of interest and spectral pre-processing on the detection of non-O157 shiga-toxin producing escherichia coli serogroups on agar media by hyperspectral imaging. Near Infrared Spectroscopy Journal. 20(4):10. DOI:1255/jnirs.1004,2012.
Interpretive Summary: Hyperspectral imaging was investigated as a method for presumptive positive screening of pathogenic non-O157 Shiga-toxin producing Escherichia coli (STEC) serotypes on agar plates. Since no fermentable carbohydrate sources for discrimination of non-O157 STEC are available, it is difficult to differentiate among non-O157 serotypes and from other flora growing on the agar plate and to select colonies for conformational testing. Hyperspectral imaging is an optical imaging technique that combines aspects of conventional imaging and vibrational spectroscopy so that data can provide two-dimensional spatial information on colony shapes and one-dimensional spectral information at every pixel in each colony under test. In this study, hyperspectral imaging was tested to discriminate each of 6 non-O157 STEC serotypes from each other (O26, O111, O45, O121, O103, and O145) grown on Rainbow Agar. Spectral libraries of pure pathogen cultures were built and classification models were developed. Classification accuracies were dependent on how the spectra used to build the libraries were selected from the colonies and the data pre-treatment used in model development. Results showed the potential of the hyperspectral imaging technique for rapid screening of food samples contaminated by STEC pathogens.
Food borne infection caused by Shiga toxin-producing Escherichia coli (STEC) is a major worldwide health concern. The best known STEC serotype is E. coli O157:H7,
which can be easily identified when cultured on sorbitol-MacConkey (SMAC) agar. Recently, six non-O157 STEC serotypes have been found to cause human illnesses. These serotypes ferment sorbital and form pink colonies; therefore SMAC agar cannot be used differentiate non-O157 serotypes from each other and other flora growing on the plate. This study investigated the potential of VNIR hyperspectral imaging and chemometrics to spectrally differentiate non-O517 STEC serotypes (O26, O45, O103, O111, O121, and O145) grown as spots on Rainbow agar media. Mahalanobis distances classifiers were developed using spectra obtained from ground truth regions-of interest (ROIs) of each serotype colony. The ROIs were selected as an open ellipse to include only the leading edge of growth and as a closed ellipse covering the entire colony. For each ROI type, the Mahalanobis distances classifiers were developed using data pre-treatments Log (1/Reflectance), first derivative and multiplicative scatter correction. Serotypes O45 and O121 showed consistently over 98% accuracy regardless of the classification algorithm used. First derivative spectra helped to increase the detection accuracies of the other serotypes. The classification accuracy for serotypes O26, O111, O103, and 0145 using the closed ellipse and open ellipse classification algorithms showed varying results from 25% to 84% and 83% to 100%, respectively. The lower accuracy using closed ellipse spectra was due to the very different spectral characteristics inside (ie., not inclusive of the open ellipse) the closed ellipse among serotypes. Practical implementation of the hyperspectral imaging methods to be used for presumptive positive screening of non-O571 STEC will be dependent the development of classification algorithms to identify the target bacteria in the presence of background flora grown on spread plates.