Skip to main content
ARS Home » Research » Publications at this Location » Publication #140739

Title: PHYSICAL AND OPTICAL PROPERTIES OF CORN INFESTED WITH ASPERGILLUS FLAVUS, FUSARIUM VERTICILLIOIDES AND OTHER MOLDS

Author
item Pearson, Thomas
item Wicklow, Donald

Submitted to: Aflatoxin Elimination Workshop Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 10/25/2002
Publication Date: 10/25/2002
Citation: PEARSON, T.C., WICKLOW, D.T. PHYSICAL AND OPTICAL PROPERTIES OF CORN INFESTED WITH ASPERGILLUS FLAVUS, FUSARIUM VERTICILLIOIDES AND OTHER MOLDS. AFLATOXIN ELIMINATION WORKSHOP PROCEEDINGS OF THE AFLATOXIN ELIMINATION WORKSHOP. 2002. ABSTRACT ONLY.

Interpretive Summary:

Technical Abstract: The goal of this project was to evaluate the physical and optical properties of whole corn kernels infected with one of a variety of molds and to identify features of these kernels that would identify the myotoxin contaminated kernels from non-contaminated kernels. Corn ears in the late milk to early dough stage of kernel maturity were inoculated with one of the following molds and harvested in the Fall: Aspergillus flavus, Aspergillus niger, Fusarium verticillioides, Fusarium graminearum, Diplodia maydis, Trichoderma viride, Acremonium zeae, Nigrospora oryzae, Penicillium funiculosum, Pencillium oxalicum Pencillium variable. Individual kernels were visually inspected and classified according to symptoms and severity of kernel infection. The following types of images were collected from all kernels: color, visible and NIR transmittance, fluorescence images under a black light, and x-ray. In addition, NIR reflectance spectra (500-1700 nm) of individual kernels was collected. All images and spectra were taken on both the germ side and opposite side of the kernels. Kernel thickness, weight and volume of individual kernels were also measured. A neural network was trained to classify kernels into their mold infection and symptom severity categories using the full NIR spectra. Classification results were greater than 95% for controls and kernels infected with other molds was generally greater than 80% and most of the misclassification errors only placed kernels into a different mold species or severity category. There are commercial high speed sorters capable of measuring reflected light from two different NIR spectral bands. A statistical method was developed to exhaustively search through the NIR spectra for the optimal two spectral bands to remove aflatoxin contaminated corn. It was found that the spectral bands centered at 725 nm and 1175 nm would detect greater than 99% of the kernels with aflatoxin greater than 100 ppb, and correctly classify 100% of the kernels with no detectable aflatoxin. Most kernels with measurable aflatoxin below 100 ppb were classified as non-contaminated. Applying the same decision rule on kernels infested with other molds indicate that kernels with slight symptoms were mostly classified as non-contaminated while kernels with a shriveled appearance would be classified as aflatoxin contaminated. Data from all the images were combined and used to classify infected kernels into their infecting mold species and disease severity category. The most important images for classification purposes were, in order of importance: blue channel of the color reflectance image, x-ray image, and NIR transmittance image, and NIR transmittance image at 780 nm, and UV fluorescence image. Using the mean pixel intensities and standard deviations from all these images approximately 85% of all kernels could be classified into the correct mold species and severity of damage category.