Assessing deoxynivalenol concentration reduction and mass loss in wheat batches using near infrared hyperspectral imaging

• NIR-hyperspectral imaging was used to predict deoxynivalenol in single wheat kernels. • A classification model was inserted in a desktop application. • The application was tested for automatic processing of NIR- hyperspectral images. • Tested batches showed 92 % (w) accepted fraction with mean 379 µg deoxynivalenol/kg. Deoxynivalenol is a mycotoxin which frequently contaminates wheat and its derived products causing a major concern. In this work a previously developed model was inserted into a prototype desktop application for automated prediction of individual wheat kernels as contaminated over or under the maximum set level in the EU. Using 42 batches to test the system it was observed that in batches contaminated over 1 mg/kg the rejected fraction always contained more deoxynivalenol (mean 8.1 mg/kg) than the accepted one (mean 0.6 mg/kg), with 92.5 % of accepted mass fraction, in average.