Submitted to: Journal of Food Protection
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/24/2024
Publication Date: 8/16/2024
Citation: Tao, F., Yao, H., Hruska, Z., Rajasekaran, K., Qin, J., Kim, M., Chao, K. 2024. Raman hyperspectral imaging as a potential tool for rapid and nondestructive identification of aflatoxin contamination in corn kernels. Journal of Food Protection. 87(9). 100335. https://doi.org/10.1016/j.jfp.2024.100335.
DOI: https://doi.org/10.1016/j.jfp.2024.100335

Interpretive Summary: Aflatoxins are a family of toxins produced by certain fungi that can contaminate corn pre-harvest and post-harvest. Aflatoxins are highly toxic, carcinogenic, and consuming foods contaminated by aflatoxins may lead to serious health consequences. Food processing techniques are not sufficient to eliminate aflatoxins from contaminated agricultural and food commodities due to their heat resistant nature. Our objective is to develop a rapid and nondestructive technology which is capable to be applied in a real-time and on-line mode for high-throughput screening aflatoxins in corn kernels. We employed the line-scan Raman hyperspectral imaging to identify aflatoxin contamination in corn kernels in this study. Corn kernels artificially inoculated with toxigenic, non-toxigenic fungi Aspergillus flavus and sterile distilled water (control) were imaged by the Raman system, and meanwhile, were analyzed for aflatoxin content. Statistical analyses results between the information extracted from Raman image and kernel aflatoxin content showed mean overall accuracies between 75-77% in discriminating aflatoxin-negative and -positive kernels. Five wavenumbers representing the major spectral differences between aflatoxin-negative and -positive kernels were identified Our results from this study indicate the potential of using Raman hyperspectral imaging for identification of aflatoxin contamination in corn kernels.

Technical Abstract: The potential of using line-scan Raman hyperspectral imaging with a 785 nm excitation line laser was investigated as a rapid and non-destructive analytical tool for determination of aflatoxin contamination in corn kernels. Nine-hundred kernels were artificially inoculated in the laboratory, with 300 kernels each inoculated with AF13 (aflatoxigenic) fungus, AF36 (non-aflatoxigenic) fungus, and sterile distilled water (control). One-hundred kernels from each treatment was subsequently incubated for 3, 5, and 8 days. The mean spectra of single kernels were extracted from the endosperm side and the embryo area on the germ side, and the local peaks were identified based upon the calculated reference spectra of aflatoxin-negative and -positive categories, separately. Discriminant models were established using the identified local peaks as inputs. The modeling results under the condition of training sample size ratios of 1:1 and 2:1 in terms of aflatoxin-negative to -positive kernels showed that overall, the discriminant models achieved more balanced prediction accuracies for the aflatoxin-negative and -positive categories with the 1:1 ratio. The models using the germ-side local peaks achieved comparable predictive capability to the models using the full peak set of both kernel sides, and both types of models performed better than the models using the endosperm-side local peaks. Based upon the full peak set of both kernel sides, the discriminant models with the 20-ppb threshold achieved mean prediction accuracies near 74%, 80-81%, and 75% in terms of aflatoxin-negative and -positive categories, and overall accuracy; and correspondingly, the models with the 100-ppb threshold achieved mean prediction accuracies of 76%, 87%, and 76-77%. The major difference between the local peak sets of the aflatoxin-negative and -positive categories were located near the wavenumbers 1687 and 1820 cm-1 over the endosperm side and 1671, 1955 and 2109 cm-1 over the germ side. The prediction accuracies of the simplified models using these 5 peaks were slightly inferior to the models using the full peak set of both kernel sides (including 60 peak variables), but the establishment of the simplified models using 5 peaks indicates a potential to develop a rapid and nondestructive aflatoxin detection technology for real-time and on-line applications.