Submitted to: American Phytopathology Society
Publication Type: Abstract Only
Publication Acceptance Date: July 28, 2006
Publication Date: July 28, 2006
Citation: Wicklow, D.T., Pearson, T.C. 2006. Detection and sorting of single symptomatic maize grains infected by different fungal species and contaminated with mycotoxins [abstract]. American Phytopathology Society. Technical Abstract: Grains highly contaminated by aflatoxin and fumonisin are unevenly distributed in a grain lot and may be concentrated in a very small percentage of the product. Near-infrared (NIR) reflectance spectra (500-1700 nm) were analyzed to select the pair of absorbance bands (filters) giving the lowest classification error rate for removing whole yellow maize grains contaminated with aflatoxin (750–1200 nm) or white maize grains contaminated with fumonisin (500-1200 nm) in a single pass through a commercial high speed sorter (@ 7000 Kg/hr). Our research also seeks to classify individual grains infected with different fungal species and to distinguish resistance and susceptibility reactions among corn varieties. Neural networks are being trained to classify grains by fungal species using principle components of the full reflectance spectra. Spectra of single maize grains can be measured automatically and grains with multiple symptoms and mycotoxins can be sorted into different fungal species categories at rates of about 1 per second using commercial instruments. Our initial work has shown that classification accuracies for severely discolored grains infected with Aspergillus flavus, Stenocarpella maydis, Fusarium graminearum, Fusarium verticillioides, and Trichoderma viride averaged 92.1% and 94.8% for two commercial corn hybrids. Protective endophytes, including mycoparasites that live asymptomatically in maize, are not readily distinguished from uninfected grains.