|SEO, YOUNG WOOK - US Department Of Agriculture (USDA)|
|Hinton, Jr, Arthur|
Submitted to: Applied Engineering in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/13/2014
Publication Date: 7/7/2014
Citation: Seo, Y., Yoon, S.C., Park, B., Hinton Jr, A., Windham, W.R., Lawrence, K.C. 2015. Development of hyperspectral imaging technique for salmonella enteritidis and typhimurium on agar plates. Applied Engineering in Agriculture. 30(3):495-506. (doi:10.13031/aea.30.10435).
Interpretive Summary: Detection and enumeration of bacteria typically require the inoculation and incubation of microorganisms on agar plates before finding and picking up the presumptive positive colonies on agar plates, which is labor-intensive and time-consuming. In addition, growth of background microflora on agar plates along with foodborne pathogens such as Salmonella and Campylobacter may also significantly affect the performance of visual screening of agar plates. In this study, visible near-infrared hyperspectral imaging measuring both spatial and spectral information in the wavelength range of 400-1,000 nm was used to detect Salmonella Enteritidis (SE) and Salmonella Typhimurium (ST) on brilliant green sulfa (BGS) and xylose lysine tergitol 4 (XLT4) agar plates and differentiate them from background microflora commonly found in chicken carcass rinses. Ten classification models for Salmonella detection in the presence of background microflora were compared by testing them on the training data obtained from pure cultures of two Salmonella serotypes (SE and ST) and eight known background microflora. The accuracy in detecting Salmonella grown on BGS agar was 98%. Salmonella Typhimurium on XLT4 agar could be detected with over 99% accuracy. Validation of the Salmonella detection models on independent test samples obtained from Salmonella-spiked chicken carcass rinses inoculated on BGS agar showed the potential of hyperspectral imaging for Salmonella detection by reaching the detection accuracy rate of about 99%. The expected outcome of the research is a rapid and accurate screening tool for the detection of Salmonella colonies on agar plates.
Technical Abstract: Salmonella is a common cause of foodborne disease resulting from the consumption of contaminated food products. Although a direct plating method is widely used for presumptive positive screening of pathogenic Salmonella colonies on agar plates, it is labor-intensive, time-consuming and also prone to human errors. This paper reports the development of a hyperspectral imaging technique for automated screening of the two common serotypes of Salmonella, Salmonella Enteritidis (SE) and Salmonella Typhimurium (ST), grown on agar plates and for differentiating them from background microflora often found in poultry carcass rinses. Visible near-infrared hyperspectral imaging, providing the spectral and spatial information unique to each microorganism, was utilized to differentiate SE and ST from the background microflora on brilliant green sulfa (BGS) and/or xylose lysine tergitol 4 (XLT4) agar plates. Five different machine learning algorithms including Mahalanobis distance (MD), k-nearest neighbor (kNN), linear discriminant analysis (LDA), quadratic discriminant analysis (QDA), and support vector machine (SVM) in addition to a multivariate data analysis method, the principal component analysis (PCA), were compared to determine the best classification algorithm in Salmonella detection and classification. When trained on the data from pure cultures of Salmonella and known background microflora, the classification accuracy of each classification algorithm in detecting Salmonella on BGS agar was about 98% on average, although it was difficult to differentiate between SE and ST. The classification accuracy in detecting Salmonella colonies on XLT4 agar was about 88% on average while the detection accuracy for ST colonies were over 99%. The validation of the classification algorithms with independent test samples of chicken carcass rinses spiked with SE and ST showed that the best performance was achieved by QDA with the prediction accuracy of about 99% (Kappa coefficient=0.97).