Submitted to: Corn Dry Milling Conference Proceedings
Publication Type: Abstract Only
Publication Acceptance Date: 6/2/2005
Publication Date: 6/2/2005
Citation: Pearson, T.C., Wicklow, D.T. 2005. Nir spectroscopy as a tool for optimizing sorting of white corn kernels contaminated with fumonisin [abstract]. 46th Annual Corn Dry Milling Conference. p. 7. Interpretive Summary:
Technical Abstract: Near infrared and reflectance spectra (500-1700nm) were analyzed to determine if they could be used to identify single whole white corn kernels contaminated with fumonisin. Kernels used for the study were obtained from processors in Illinois, Indiana, Kentucky, and Nebraska. Kernels were visually examined and grouped into six symptom categories: asymptomatic, chalky tip end, yellow-tan tip end, red streaks, 50% discolored, and 100% discolored. Friable kernels and fragments were not included in this study as they are usually removed by existing cleaning equipment at grain elevators. Spectra were acquired on both the germ side and endosperms side of each kernel. After spectra acquisition, kernels were weighed individually and then placed in groups of five according to their classification based upon symptoms of fungal infection and numerical sequence within each pill box. Total fumonisin (Bl, B2, and B3) was measured with a fluorometer after extracts were purified with immunoaffinity columns (Fumonitest, Vicam, Watertown, MA) using the procedure recommended for corn, sorghum, and 17% protein poultry feed. The fumonisin level of each five-kernel group then was assigned to each individual kernel from that group. Kernels were analyzed in groups instead of individually to reduce cost and analysis time. Mycological evaluations, performed on grain sub-sampled from each symptom category and state, revealed that the five kernel groupings risk producing false positives. For high speed sorting operations, whole spectra cannot be acquired at throughput rates that are economically feasible. Most commercial sorting machines are able to only measure one spectral band of light while some machines can measure two bands. Discriminate analysis was used to select the optimal pair of wavelengths to identify kernels containing aflatoxin. It was found that using the wavelength pair of 500nm and l200nm, approximately 77% of the kernels having high levels of fumonisin (>40ppm) were correctly classified. Additionally, approximately 96% of the kernels having low levels offumonisn <2 ppm) were correctly classified. In contrast, if only a single band is selected for distinguishing contaminated kernels, the accuracy for kernels having low fumonisin levels <2 ppm) drops to approximately 83%. Thus, use of a dual band sorting machine for removal of white corn contaminated with fumonisin would result in 13% less good product being removed than with a monochromatic sorter. Previous work with yellow corn showed that approximately 85% of the aflatoxin and fumonisin could be removed by high speed sorters using the spectral bands of 750nm and l200nm. It was hypothesized that the 750nm band was detecting some color changes in fungal infested kernels while the l200nm band was responding to increased porosity of the degraded endosperm. Insect damaged kernels have low absorbance at l200nm, due to feeding and fungal infestation, and would all be rejected. In the case of white corn, 500nm was found to be more accurate than 750nm for the visible spectral band. This may be due to the white corn germ and endosperm being of more uniform color than yellow corn kernels with a white germ. Because yellow corn absorbs more light at 500nm, asymptomatic yellow corn kernels can be distinguished from white corn kernels.