The objective of this research was to develop a neural network based method for determination of Deoxynivalenol
(DON) levels in barley using near–infrared (NIR) spectroscopy. The NIR spectra of 188 barley samples with DON level
between 0.3 to 50.8 ppm were collected using the FOSS NIR Systems 6500 Near Infrared Spectrometer. The DON levels were
measured with GC/mass spectrometry (GC/MS). With the NIR spectra as input variable and GC/mass measured DON as
output variable, neural networks were developed and trained to predict DON levels. The prediction accuracy of the models
was tested using randomly selected production sets, which had not been seen by the models. The effects of wavelength interval
and ranges on the prediction accuracy of models were also examined. The results demonstrate that the combination of neural
networks and NIR spectra can be conveniently used to determine DON levels in barley.