|
Aflatoxins are among the most carcinogenic mycotoxins and are known to contaminate a wide variety of agricultural and food commodities. This study aims to explore the effectiveness of Raman hyperspectral imaging in detecting aflatoxin contamination in corn kernels in a rapid and non-destructive manner. Four hundred kernels were used with 2 treatments, namely, 200 kernels inoculated with the AF13 fungus (aflatoxigenic), and 200 kernels inoculated with sterile distilled water as control. One hundred kernels from each treatment were incubated at 30 °C for 5 and 8 days, separately. On the specified post-inoculation day, the kernels were dried and wiped free of surface mold prior to imaging. The Raman images of kernels were acquired over the endosperm side over the 103-2831 cm-1 wavenumber range. The standard aflatoxin concentration in each kernel was determined by the VICAM AflaTest method. The original mean spectra of single kernels were extracted and preprocessed by adaptive iteratively reweighted penalized least squares, Savitzky Golay smoothing and min-max normalization. On basis of the calculated “reference” mean spectra of the aflatoxin negative and -positive categories, 14 and 17 local peaks were determined, separately. After removing the identical peaks from both peak sets, a total of 24 unique peaks were extracted and used as inputs for further discriminant model development. With 20 ppb and 100 ppb as the classification thresholds, the 2-class discriminant models established with the principal component analysis-linear discriminant analysis and partial least-squares discriminant analysis methods, obtained mean overall prediction accuracies between 77.9% and 82.0%. Further investigation is ongoing to include more diverse samples and execute different types of computation algorithms, seeking solutions to improve the discriminant models in identifying aflatoxin contamination in corn kernels.
|
