Researchers (Prof Marena Manley & Dr Paul J Williams)
at the Department of Food Science, Stellenbosch University, have been using this technology for some time now
to investigate its efficacy for use as a tool for food quality and safety. Dr Paul James Williams recently completed
his PhD (2013) using this method to detect fungal contaminated maize kernels, differentiate between fungi
commonly associated with maize infection and study
fungal growth. In his work, detecting the cereal killer
(Fusarium verticillioides) was of paramount importance.
This culprit is responsible for major losses in the grain
industry, more importantly it is associated with the production of mycotoxins that are harmful to humans and
animals. With current grading practices, infected grain
often enters the food chain undetected due to the subjectivity and laboriousness of grading techniques, making NIR hyperspectral imaging the ideal candidate for
objective, rapid sorting. It was shown that 20 h after
inoculation with F. verticillioides, infected maize kernels
could be distinguished from uninfected kernels (P. J. Williams, P. Geladi, T. J. Britz, & M. Manley, 2012). Furthermore, it was possible to differentiate between closely
related Fusarium species (P. Williams, P. Geladi, T. Britz,
& M. Manley, 2012b), and study their growth on agar
(a)
(b)
media (P. Williams, P. Geladi, T. Britz, & M. Manley,
2012a). Not only was it possible to distinguish between
the “growth rings” of the fungal colonies, but it was
also possible to construct profiles resembling growth
curves.
Currently an MSc in Food Science student is using the
technique to develop classification models that will be
capable of grading maize kernels according to regulations stipulated in South African legislation (Act No, 119
of 1990, as amended by Government Notice No. R. 473
of 8 May 2009). Multispectral imaging (similar to hyperspectral imaging just with fewer wavelengths) and NIR
hyperspectral imaging are being evaluated. The kernels
studied were divided into 13 groups based on quality by
qualified maize graders, and were imaged with both a
SisuChema NIR hyperspectral push broom imaging system (Specim, Spectral Imaging Ltd, Oulu, Finland), and a
Video meter multispectral imaging system (Videometer
A/S, Hørsholm, Denmark). The calibrationcoefficients for
the multispectral data ranged 0.46 to 0.95, with correct
classifications ranging 83 to 100%. The calibration coefficients for the NIR hyperspectral data ranged 0.65 to
0.88, with correct classifications ranging 69 to 100%.
(c)
Figure 2. Collection of images illustrating the difference between a chemical image (a), a digital image
(b) and a diagram (c). The information in (c) was used to categorize a hyperspectral image to obtain (a),
a classified image where each colour represents a different class.
These are but a few of the applications currently under
investigation, others include detection and differentiation of foodborne pathogens on growth media, distinction of Fusarium species on growth media and shedding light on food fraud. In this manner, researchers at
SU are making the invisible, visible.
References
Williams, P., Geladi, P., Britz, T., & Manley, M. (2012a).
Growth characteristics of three Fusarium species evaluated by near-infrared hyperspectral imaging and multivariate image analysis. Applied Microbiology and Biotechnology, 96(3), 803-813.
Williams, P., Geladi, P., Britz, T., & Manley, M. (2012b).
Near-infrared (NIR) hyperspectral imaging and multivariate image analysis to study growth characteristics and
differences between species and strains of members of
the genus Fusarium. Analytical and Bioanalytical Chemistry, 404(6), 1759-1769.
Williams, P. J., Geladi, P., Britz, T. J., & Manley, M.
(2012). Investigation of fungal development in maize
kernels using NIR hyperspectral imaging and multivariate data analysis. Journal of Cereal Science, 55(3), 272278.